The Impact of the Introduction of Multicore Technologies on the Computing Market and Opportunities for Europe FINAL REPORT

Transcription

1 FR The Impact of the Introduction of Multicore Technologies on the Computing Market and Opportunities for Europe FINAL REPORT A study prepared for the European Commission DG Communications Networks, Content & Technology

2 This study was carried out for the European Commission by Authors: Gabriella Cattaneo, Chris Ingle, Martin Canning, with Stefania Aguzzi, Steve Conway, Mario Morales, Varun Srikumar Internal identification Contract number: 30-CE /00-34 SMART number: 2011/0066 DISCLAIMER By the European Commission, Directorate-General of Communications Networks, Content & Technology. The information and views set out in this publication are those of the author(s) and do not necessarily reflect the official opinion of the Commission. The Commission does not guarantee the accuracy of the data included in this study. Neither the Commission nor any person acting on the Commission s behalf may be held responsible for the use which may be made of the information contained therein. ISBN: DOI: /10157 European Union, All rights reserved. Certain parts are licensed under conditions to the EU. Reproduction is authorised provided the source is acknowledged. 2

3 Table of Contents EXECUTIVE SUMMARY Introduction and Background Background Landscape and Scope Structure of the Report Multicore technologies: trends and impacts Overview Semiconductor Industry Analysis Multicore End-User Industry Analysis Embedded Systems Industrial Sector Summary General-Purpose and High-Performance Computing General-Purpose Server Market Final Considerations on the SWOT Analysis Main Challenges and Areas of Action Overview Developing a Vertical Integration Network Model Description of the Challenge Possible Actions Example Structure Potential Impacts Developing Next-Generation Fab Capability in Europe Description of the Challenge Possible Actions Potential Impacts Building on the Low Power Ecosystem Description of the Challenge Possible Action Potential Impact Development of Heterogeneous Hardware/Software Architectures Description of the Challenge Possible Action Potential Impact Development of new Software Tools, Systems and Applications Suited for Parallelism Description of the Challenge Possible Action Potential Impact Remove Barriers to Growth for High-Tech SMEs Description of the Challenge Possible Action

4 Potential Impact The socioeconomic relevance of multicore innovation Overview The Employment Footprint of the Multicore Ecosystem in Europe The Role of ICT Component Manufacturing Clusters Meeting Social Challenges Conclusions and Recommendations Overview Recommendations Methodology Annex Table of Figures Figure 1 Semiconductor Value Chain, 2011 ( B) Figure 2 Overview of the Market and Technology Challenges Figure 3 Worldwide Overall Semiconductor Revenue Forecast ( B) Figure 4 Worldwide Semiconductor Revenue by Major Market Segment, 2011 ( 227B) Figure 5 Comparative Analysis of Multicore End User Industries Growth Potential Figure 6 Analysis of Strength and Growth of Multicore End User Industries Figure 7 Server Market Revenues in the EU27 Compared to Rest of EMEA, Figure 8 Server Market Revenues in the EU27, Historical 2011 and Forecast , Millions and Shipments Figure 9 Main Research and Market Challenges Figure 10 Approach to the Development of Vertical Integration Models Figure 11 Performance per Watt Comparison for Caldexa CPU and Intel CPU Figure 12 Multicore Industry Employment Footprint in the EU, Figure 13 Multicore User Sectors Employment in the EU, Figure 14 Mapping of ICT Components Manufacturing Clusters in the EU Table of Tables Table 1 Competitive Positioning of Main EU Semiconductor Companies Table 2 SWOT - European Automotive, Industrial Automation, Energy, Healthcare, Aerospace Table 3 Consumer Electronics SWOT Analysis (I) Table 4 Consumer Electronics SWOT Analysis (II) Table 5 Communication and Wireless Infrastructure SWOT Analysis Table 6 Worldwide ESD Packaged Software Revenue, ( M) Table 7 Employment in Electronic Component Manufacturing by MS, Table 8 Employment in Electronic Component Manufacturing by MS, Table 9 Multicore Innovation meeting Social Challenges in H Table 10 List of Companies Interviewed Table 11 IDC Taxonomy of Semiconductor Market Segments

5 E X E C U T I V E S U M M A R Y Drivers of Change The increasing diffusion of multicore technologies is profoundly changing the semiconductor industry and affecting the dynamics of all multiprocessor and system-on-a-chip (SoC) enduser markets. This transition is driven by technology innovation. The semiconductor industry is moving away from increasing performance through clock speed increases and toward increasing performance through building more capability for a given space of silicon. The performance gains from increasing clock speed come with an associated increase in power consumption which provides a challenge across the computing ecosystem both in terms of cost of power and in terms of dissipating heat in both general-purpose and embedded systems. To address this problem the industry is introducing a combination of new techniques, not only multicore/manycore technologies but also SoC and heterogeneous systems. According to IDC estimates, these technologies will take over the semiconductor market by Multicore processor revenues in the EU will more than double between 2011 and 2020, growing from 5.4 billion to 12.7 billion in This covers the consumer electronics, communication and wireless, automotive, energy, industrial automation, and healthcare sectors. Consumer electronics and communication and wireless represent the lion's share of this market and they already have a high penetration of multicore processors, with moderate growth foreseen to The EU industry has a weak positioning in these segments (in decline from former strength in communication), even if the absolute size of these markets makes them attractive. According to IDC, the opportunities for growth for the EU industry are stronger in the other market segments: first of all in the automotive and energy sectors, with medium size but high attractiveness, but also in industrial automation and healthcare, somewhat smaller but potentially attractive. It is not surprising that these last segments fall within the broad area of embedded systems, where the EU has traditional expertise. In all these industries, the adoption of multicore is not simply an incremental innovation (improving speed and power): it enables a new level of process integration and innovation, accompanied by sophisticated applications and services, with advanced control and management capabilities. The comparative analysis of the EU industry SWOT confirms that the window of opportunity opened by the multicore innovation is not limited to a collection of market niches; rather it cuts across the most important sectors of the European economy. It is therefore relevant and useful to identify the horizontal technology and market challenges to be overcome to enable the EU industry to exploit the multicore opportunity. Socioeconomic Relevance of Multicore Innovation A successful migration to multicore has implications not only for the competitiveness and strength of the industry, but for the overall development of the European socio-economic system. Conversely, the inability to deal with the research and technology challenges bears potential risks for the EU economy, in terms of potential jobs loss and/or lack of employment growth, and of missing innovation in a wide range of user sectors. The multicore value chain extends from semiconductor manufacturing to the embedded systems markets. By adding up the number of employees in all the industries comprised by this ecosystem we develop an estimate of the employment "footprint" of the multicore ecosystem. The employment "footprint" of the multicore industry in Europe covered 2.9 million jobs in 2010, plus 8 million jobs in the multicore user industries, for a total of approximately million jobs (source: Eurostat 2010). This corresponds roughly to 20% of the total 5

6 employment in industry sectors in Europe, split between 5% for the supplier industries and 15% in the user industries. This shows that the multicore ecosystem has a high relevance for the European economy and labor market. The new products and services enabled by multicore technologies will contribute to respond to the main social challenges identified by Horizon More specifically, multicore innovation will contribute to the development and deployment of connected cars, automated driving (improving safety and ease of driving for the elderly and the disabled); of smart grids, smart homes, smart buildings (reducing energy costs and consumption, as well as minimizing the carbon footprint thanks to greater use of renewable energies); of more advanced industrial automation (improving safety and work conditions in factories, increasing productivity); of automated POS and all kinds of new services in retail, based on mobile connectivity integrated with social networks, improving the consumers' experience. Main Challenges The phase of development of the microprocessor and multicore solutions market offers an extremely interesting but challenging opportunity to the EU industry and economy. The first opportunity is to strengthen and grow the fragmented EU industry into an open and thriving supply ecosystem, able to support next generation system requirements, characterized by a flexible leading semiconductor manufacturing network, silicon integration capabilities, software tool development capabilities; an EU industry which should not be afraid to compete globally. A key aspect is to close the gap between hardware and software advances. Although multi-core processors have been mainstream for several years, current software development still lags behind, with sequential applications still the dominant program design. According to IDC, only fundamental redesign and development of new software can enable the exploitation of new generation hardware in all fields, not only HPC. For the EU software industry, this is an important opportunity to gain back lost ground and position itself more strongly for the next decade of innovation. But equally important is to insure that the EU end user sectors are able to access and use at best multicore technologies and solutions appropriate for their business, in order to develop their competitiveness. These two opportunities are deeply interconnected, since the ability to develop and exploit the right technologies builds on an open and productive demand-supply interaction. To achieve these goals, there are 6 critical research and market development challenges which, in our view, are the most relevant. For each of these main challenges we have identified the possible actions by the EC as follows. Recommendations Business Models Innovation and Reorganization of the Value Chain The European Commission should promote the development of vertical integration in the semiconductor and embedded systems industries and with the, by promoting the creation of industry working groups and collaborative networks with the goal to develop common requirements, common reference architectures, common standards, and common code base. Research and Innovation in Horizon 2020 Concerning Horizon 2020, the EC should take full advantage of the new program's extended focus on research and innovation to promote the development of multicore technologies focusing on the following research challenges: 6

7 Development of cost-effective, advanced software tools and systems suited for parallelism, enabling the development of new software and applications, and/or the reengineering of legacy codes. Research into the development of heterogeneous hardware and software architectures that are aligned with requirements in key industries such as automotive, healthcare, and energy while avoiding the trap of an over-proliferation of architectures, including safety and mixed criticality, as well as the development of cross-component/cross-layer optimization for design integration. Research into advanced software requirements, deriving efficiencies from parallelism in the code, software partitioning, running different operating systems simultaneously, and associated application software geared towards data or control plane functions. Organization of Research Promote a system of research centers of excellence (rather than networks) differentiated by specialization (or domain) to focus on Europe's areas of strength. Promote research and development projects focused on the development of precommercial new applications and services based on multicore innovation, which could address the vertical markets targeted by embedded systems. In terms of developing new reference architectures, it is likely that existing architectures already provide a robust platform for future-oriented enhancements. A strong focus on commonality and reusability should be leading characteristics of research initiatives. In terms of crossing the "valley of death" between research and market exploitation, European vendors and technology suppliers should be encouraged to become first movers in the migration toward heterogeneous architectures and systems across the full value chain. Define a clear research roadmap to achieve the transition to parallel software, including the timescale of objectives, pursuing both the adaptation of existing software and the research on new programming paradigms. The development of systems assisting with parallelizing existing software and compiler tools is a short-term objective; the development of new programming paradigms and self-adapting software with full performance portability is a long term objective. Developing Pilot Lines and Research in Cutting-Edge Manufacturing Technologies in Europe In order to maintain competitiveness of the EU industry and carefully manage the intellectual property at each stage of the multicore/manycore value chain, sustaining the competitive positioning of fabless operators, the EC should: Increase the investments in device fabrication pilot lines and prototypes to produce limited quantities of devices at reasonable cost, to be accessible for high-tech SMEs and any other European player interested in the design and development of new multicore/manycore. Give careful consideration as to whether a pilot-line initiative should strive to stay at the leading edge of semiconductor technology or aim to focus on older process capabilities that will still provide a robust foundation for product testing and proving. Align strategic decisions with the broader strategy around the future of the European semiconductor industry and the industry transition to 450mm wafers. Carry out research on how potentially disruptive new technologies/materials such as Graphene will change the semiconductor landscape and impact the European industry. This is potentially a separate research program which is relevant to the multicore-manycore topic specifically, but has implications for the full semiconductor industry in Europe within the context of horizon 2020 and beyond. 7

8 Building on the Low Power Ecosystem There is one leading EU player in the low power technologies field ARM which has been able to develop a worldwide business ecosystem. The EC should look for ways to build on this ecosystem in order to multiply the potential competitive advantages for the whole EU industry and expand the value added for Europe. To do so the EC should, within Horizon 2020: Carry out research on the governance needed and the challenges created by low power designs, which are being developed in the U.S. and Asia, primarily by existing vendors located in the same world regions. However this disruption creates opportunities for start-ups in Europe to build a place in the market. Support the recompilation of software needed to run in the software ecosystem for ARM, which will develop in the next few years. Regular, off-the-shelf Windows and Linux software and applications cannot run on ARM processors and need recompiling, tuning and optimizing. Promote the development of a Linux-based software ecosystem revolving around ARM technologies and support the creation and growth of start-ups that can build applications to run on top of the platform. Remove Barriers to Growth for High-Tech SMEs In order to remove barriers to growth for high-tech SMEs the EC should: Promote the development of start-ups and spin-offs in advanced multicore technologies: o Leverage the presence of strong ICT components industry clusters in Europe to provide support to start ups or small high tech companies through their infrastructures. o Encourage the development of smart coaching networks, learning communities, and innovative forms of "crowdfunding" for seed funding new ideas and help to train the would-be entrepreneurs in the industry. Make sure that start-ups and spin-offs have access to the latest advanced silicon technologies at affordable prices, including IDM production lines. Build the bridge between research and market opportunities. Develop awareness campaigns on multicore benefits and promote the development of business and use cases of adoption of innovative services based on multi-manycore. Continue the initiatives started in FP Work Programme through the Factory of the Future PPP, to develop and test simulation services for engineering/manufacturing SMEs over HPC-Clouds. Support and enable the pre-commercial procurement of prototype systems, to bring research closer to the problems of exploitation. This should also help start-ups or hightech SMEs to get reference customers and gain a commercial foothold in the market. Promote Education and Skills in Parallel Programming and Multicore Technologies In the field of R&D&I policies, contribute to the development of awareness raising and communication initiatives promoting microelectronics education and careers as being attractive, prestigious and a future "job for life" with both personal and financial reward. The EC should cooperate with industry and national government to increase industry based training and certification (IBTC) offers for up-skilling ICT practitioners but also for re-training side-entries, in order to respond to the emerging demand for parallel programming and other e-skills necessary to the multicore market and industry. Start an initiative for the development of European guidelines and quality labels for new curricula in the parallel programming domain. A common theme running through these recommendations is the need to overcome the fragmentation of the EU industry and market, focusing investments and efforts in selected priority areas where there is a chance to make a real difference for European 8

9 competitiveness. Our recommendations therefore suggest building on collaborative and coordinated efforts, looking for economies of scale and scope. Horizon 2020 does seem an appropriate instrument to respond to these challenges. Another critical issue is that the absolute level of investment in this field should be raised. Even before the economic crisis, the European industry has underinvested for years in the semiconductor industry. However, without considerable investments in next generation hardware and software systems the EU semiconductor and embedded systems industry risks to decline into irrelevance, exposing the EU user sectors to the risks of dependence from global suppliers. On the other hand, the good news is that the potential for success exists, if Europe will accept the challenge to build on its existing strengths and competences. Background This is the Final Report of the study "Impact of the introduction of multicore technologies on the computing market and opportunities for Europe," assigned by the European Commission, DG Connect to IDC. The overall objective of this study is to support the Horizon 2020 strategy in the computing area through a thorough assessment of the changes to the hardware and software market that are driven by the current use, and predicted growth, of multicore and heterogeneous system on chip processors. The recommendations presented have been discussed and validated in a workshop with the main stakeholders on November 28, We wish to give special thanks to all the workshop participants and to the following experts who provided valuable feedback for this report: Dr Marc Duranton: Responsible for Roadmap, HIPEAC Network, CEA Nano-INNOV, Embedded Computing Laboratory Dr Jean-Luc Dormoy: IT and Energy Consultant Prof Dr-Ing Rolf Ernst: Institut für Datentechnik und Kommunikationsnetze, Braunschweig University John Goodacre: Director, Program Management, CPU Group, ARM Alun Foster: Program Manager, Artemis Joint Undertaking Laurent Julliard: Directeur Solutions & Services Logiciels, Kalray Lutz Schubert: Head of Department HLRS, High-Performing Computer Centre, Stuttgart, University of Stuttgart 9

10 10

11 1. I N T R O D U C T I O N A N D B A C K G R O U N D 1.1. Background This is the Final Report of the study "Impact of the introduction of multicore technologies on the computing market and opportunities for Europe" carried out by IDC on behalf of the European Commission, DG Connect. The overall objective of this study is to support the Horizon 2020 strategy in the computing area through a thorough assessment of the changes to the hardware and software market that are driven by the current use, and predicted growth, of multicore and heterogeneous system on chip processors. The main goal of this report was to analyze in detail the actual and potential multicore market in Europe, for the period focusing on the following issues: Analysis of the areas of strength and weakness for Europe compared with the U.S. and Asia in respect of the computing ecosystem. Identification of the areas of the software and hardware ecosystem where Europe has opportunity to specialize and grow with the associated benefits to the European economy and society. Identification of the areas of the software and hardware ecosystem where intervention at a European level will enable Europe to achieve a higher level of economic and social benefit. The report presents the main conclusions and recommendations developed on the basis of the research conducted in the period March October 2012, leveraging IDC's extensive databases and models of the semiconductor market for the forecast model, interviewing 31 leading stakeholder in Europe, the U.S. and Asia, and organizing a workshop with 30 participants in May The conclusions and recommendations have been discussed and validated in a workshop with the main stakeholders on November 28, 2012, and the report has been revised accordingly. We wish to give special thanks to all the workshop participants and to the following experts who provided valuable feedback for this report: Dr Marc Duranton: Responsible for Roadmap, HIPEAC Network, CEA Nano-INNOV, Embedded Computing Laboratory Dr Jean-Luc Dormoy: IT and Energy Consultant Prof Dr-Ing Rolf Ernst: Institut für Datentechnik und Kommunikationsnetze, Braunschweig University John Goodacre: Director, Program Management, CPU Group, ARM Alun Foster: Program Manager, Artemis Joint Undertaking Laurent Julliard: Directeur Solutions & Services Logiciels, Kalray Lutz Schubert: Head of Department HLRS, High-performing Computer Centre, Stuttgart, University of Stuttgart 1.2. Landscape and Scope This report is motivated by a move within the semiconductor industry away from increasing performance through clock speed increases and toward increasing performance through building more capability for a given space of silicon. The performance gains from increasing clock speed come with an associated increase in power consumption which provides a 11

12 challenge across the computing ecosystem both in terms of cost of power and in terms of dissipating heat in both general purpose and embedded systems. Moreover, the concepts behind "fat" or ever more complex processors are stretched to their limits: new concepts have to come into play to translate more transistors ("freely" provided by Moore's law) into more computing power. To summarize, multi-manycores are a disruption in computing platform design in terms of computing power as well as energy consumption. To address the problem of increasing power consumption, semiconductor designs are moving to use a range of approaches which provide improvements in performance per watt of energy consumed. There are various techniques for achieving this and a number of names used to describe those techniques including: Multicore the use of two or more cores on a single die; System on chip (SoC) multiple processor functions on a single processor; Heterogeneous systems multiple computational units in a single system. This report takes two perspectives on the impact of these developments. Firstly, we look at the industry involved in bringing multicore systems to market. This includes a range of activities such as designing and fabricating processors and upstream activities such as manufacturing fabrication equipment, through to writing software to take advantage of the capabilities of these processors. For the sake of convenience we use the term "semiconductor ecosystem" in this report to encompass all these activities, this should be taken to include hardware elements, software and associated services (such as porting code from older architectures to newer architectures). Secondly, we look at the needs of users of the products of the semiconductor ecosystem. Broadly this can be split into three groups: general purpose computing such as PCs and servers; high performance computing; embedded systems, where semiconductors are part of an industrial or consumer product. The majority of this report addresses the embedded systems market however we comment on other markets where appropriate. The figure below gives a visualization of the activities involved and the global size of each activity in Figure 1 Semiconductor Value Chain, 2011 ( B) Note: IDC 2011 Annual exchange rate of Euros per dollar is assumed for the forecast period. 12

13 Develop a Vertical Integration Network Model Source: IDC, 2012 We have used IDC's taxonomy of industries within the embedded systems market. These are defined as illustrated in Table 10 presented in the annex Structure of the Report The report is structured as follows: Chapter 1 provides a short introduction to the aim of the report and defines the scope and key concepts used in the study. Chapter 2 provides an analysis of the semiconductor ecosystem and the embedded systems market in Europe, based on the research carried out in the first part of this study. This section compares Europe with other world regions and provides a SWOT analysis of the European embedded systems market. Chapter 3 presents the main research and competitiveness challenges faced by the EU semiconductor ecosystem and the embedded systems industry, in order to implement and exploit multicore-based innovation. The figure below presents a summary view of these main challenges, in a simplified "stack" reflecting the structure of the value chain of the multicore ecosystem, building from the manufacturing challenges (bottom layer) to the market policy challenges (top layer). One challenge, possibly the most critical for the EU system, cuts across the layers and consists in the need to develop innovative strategic business models based on vertical integration platforms. The analysis of challenges includes the possible actions and the potential impacts. Chapter 4 analyzes the potential socio-economic impacts of multicore technologies and the way they can meet the social challenges identified by Horizon Chapter 5 presents the main conclusions of the study and the main policy recommendations for EC actions to solve the challenges identified, discussed and validated at the workshop organized with the main stakeholders on November 28, The methodology annexes include IDC's research methodology for the analysis of the multicore market and the description of the data collection and survey work undertaken for this project. Figure 2 Overview of the Market and Technology Challenges Remove barriers to growth for high-tech SMEs Start-Ups and Spin-offs Enable the development of new software tools and new software applications for Parallelism Develop heterogeneous hardware and hoftware architectures Build on the Low Power Ecosystem Develop Next Generation Fab Capability in Europe Source: IDC

14 2. M U L T I C O R E T E C H N O L O G I E S : T R E N D S A N D I M P A C T S 2.1. Overview This chapter provides an overview of the semiconductor industry and the use of semiconductors by the different industrial segments. The section on semiconductor industry summarizes the market size and growth opportunity for the industrial segments within the EU region. Further, the industry sector analysis section describes the industrial segments in detail and the implications of changes within the market Semiconductor Industry Analysis The increasing diffusion of multicore technologies is changing profoundly the semiconductor market and affecting the dynamics of all the end-user markets consuming multiprocessors and SOCs. According to IDC estimates, multicore processors revenues in the EU will more than double in the period , growing from 5.4 billion to 12.7 billion in This corresponds to multicore revenues only in the consumer electronics, communication and wireless, automotive, energy, industrial automation and healthcare sectors. Of these, consumer electronics and communication and wireless already present a high share of penetration of multicore technologies, with a moderate rate of growth to In three other segments, automotive, energy and industrial automation, multicore technologies are only starting to be adopted, but the growth potential is extremely interesting. The specific multicore technologies revenues in the healthcare sector are likely to remain very small, compared to the other sectors. The following figures illustrate the overall industry size and the representative European industry strengths. Figure 3 Worldwide Overall Semiconductor Revenue Forecast ( B) Note: IDC 2011 Annual exchange rate of Euros per dollar is assumed for the forecast period. Source: IDC,

15 Figure 4 Worldwide Semiconductor Revenue by Major Market Segment, 2011 ( 227B) Note: IDC 2011 Annual exchange rate of Euros per dollar is assumed for the forecast period. The definition of each of the market segments is in Table 10 in the Annex. Source: IDC, 2012 These are the overall key findings within the semiconductor industry: The industries with the greatest opportunity of growth of multicore revenues in the European region are automotive and energy. There are however relevant challenges. In the automotive sector, a challenge is that most semiconductor manufacturing edge has shifted to Asia, except for Global Foundries. Furthermore, little direct collaboration exists among EU players in the automotive value chain, especially in next-generation automobile electronics ecosystems development. On the other hand Europe has relevant strengths affecting market development. Players such as ST Microelectronics and CSR have strong capabilities in sensors and power management particularly relevant for automotive. In addition, Europe has strengths in areas such as near field communication (NFC) with NXP and with ASML in optical lithography The computing and consumer sector represent the largest revenue by size in European region. In the consumer sector, personal computers and mobile devices account for approximately 50% of the global semiconductor market in terms of consumption, where Europe is less present. Europe is strongest in mobile infrastructure, automotive and industrial market segments which account for less than 20% of the global market for semiconductors. This creates a challenge of scale and alignment with the global market which must be taken into account when reflecting on how best to improve European competitiveness in this industry. This is a very concentrated industry. The top 25 semiconductor companies dominate the semiconductor industry as they account for more than 70% of overall semiconductor revenues. Rounding out the top 5 chip suppliers in 2011 were Texas Instruments, Toshiba, and Renesas Electronics, the latter two benefiting from the strong yen relative to the U.S. dollar. The next five suppliers were Qualcomm, Hynix, STMicroelectronics, Micron, and Broadcom. Together, the top 10 vendors represented 53% of total worldwide semiconductor revenues in 2011, an increase of 3% over Notable small to medium-sized companies experiencing strong growth in 2011 were Spreadtrum Communications, CoreLogic, Microsemi, Sequans, Icera, MegaChips, Nichia Chemical, Osram, RobertBosch, Skyworks, and Cavium. All these companies benefited from strong growth in the wired and wireless communications, consumer, automotive, and industrial market segments. What is important to note is that European semiconductor companies 15

16 grew closer to the rate of the overall semiconductor market which means that they have shed market share over the past couple of years. Europe only has two indigenous semiconductor companies in the top 15 global semiconductor companies by revenue. These two companies combined account for approximately 5% of the global market for semiconductors. The market trends continue to point to a closer alignment in the supply chain between semiconductor fabrication with design and software. This view should be evaluated to consider how Europe could build a more competitive semiconductor-related industry by leveraging the strong links between hardware and software, particularly in ecosystems such as Automotive and Industrial where the technology and system vendors are well positioned in Europe like Bosch, BMW, VW, Continental, Siemens, Philips, and ABB. The following table presents a brief description of the competitive positioning of the leading EU semiconductor industries. Table 1 Competitive Positioning of Main EU Semiconductor Companies Firms ARM Wolfson CSR Imaginati on STMicroel ectronics NXP Infineon ASML Brief Description of Competitive Positioning Leading position in EU microprocessor market via customers such as TI, Samsung, et al. One of the opportunities with ARM is the move to server market with partnership with companies such as HP, IBM, Dell, AMD, Nvidia. ARM is also present in the mobile phone market, where ARM's architecture is dominant (over 95% of all mobile phones) and accounts for 62% of all ARM-based volumes. A global leader in the supply of high performance mixed-signal semiconductors to the consumer electronics market. Wolfson has gained expertise in the audio components space with global mobile device companies. Opportunity for Wolfson remains in the mobile technologies coupled with the surge in tablet devices. CSR is a leading organization in design and development of silicon and software for the consumer electronics market. A pioneer in the Bluetooth solutions in the mobile phone market, CSR is now brining Bluetooth products to the automobile market with solution for in-vehicle infotainment (IVI) systems. Imagination Technologies is a leading supplier of GPU IP to tablets and smartphones, including products from Apple. With the growing interest and opportunity in automobile entertainment systems, graphics cores by Imagination (EU-based companies) are now being extensively used in the automotive IVI subsystems. A global semiconductor leader serving customers across the spectrum of electronics applications and a leading supplier of semiconductors for cellular devices. Designers of smart, connected embedded electronics for applications such as utility metering, renewable energy, and healthcare have begun adopting new class-leading microcontrollers from STMicroelectronics. NXP is a recognized leader in near filed communication (NFC) identity applications leveraging its leading RF, analog, power management, interface, security and digital processing expertise. These innovations are used in a wide range of automotive, identification, wireless infrastructure, lighting, industrial, mobile, consumer and computing applications. The future opportunity for NXP lies in the adoption and interest in electronic wallets couple with the explosive growth of mobile hardware devices. Infineon makes chips for automotive, energy efficiency and security applications. Infineon Technologies high-performance communication interface for NFC applications is gaining wide acceptance as a de facto industry standard. The importance of various NFC applications including mobile payment or ticketing in public transport is gaining momentum. ASML is Europe's largest semiconductor-equipment supplier. With the acquisition of Cymer in extreme ultraviolet lithography (EUV) technology, ASML is looking to speed up development of light sources used in ASML products. EUV technology is important for the chip-making industry to enable the production of smaller chips while increasing their capacity and speed for devices such as handsets and tablet computers. Source: IDC,

17 2.2. Multicore End-User Industry Analysis The following paragraphs will examine the main industry sectors, their trends, SWOT and the competitive positioning of the EU industry up to Table 2 SWOT - European Automotive, Industrial Automation, Energy, Healthcare, Aerospace Strengths Presence in the EU of auto OEM leaders such as BMW, Volkswagen, Saab, and of leading tier 1 suppliers such as Bosch and Continental. Research capability: EU academia and industry laboratories have been leaders in researching multicore processors, heterogeneous architectures and associated operating systems. EU electronics communication companies such as Philips, Alcatel-Lucent have deep knowledge on consumer electronics and home networking expectations. Deep innovation capabilities in healthcare. In the medical devices market, companies such as Covidien have deep expertise in imaging solutions and strong partnerships with large multinational organizations such as GE Healthcare. Opportunities Multicore adoption in nascent stage for industries such as automobiles and industrial automation, therefore providing the opportunity to lead technological developments. Global Foundries is now becoming a leader in leadingedge semi manufacturing. EU companies could benefit from its strengths and capabilities to bring semi manufacturing leadership back to the EU. M2M Communications is at the beginning stages of market adoption and allows EU companies an opportunity to establish leadership. Next-gen technologies are at adoption stage building automation, smart cities, connected car, ehealth. Weakness Semiconductor manufacturing has been in decline in the EU, and leading edge manufacturing has shifted to Asia, except for Global Foundries. Little direct collaboration in the industrial value chain & automobile value chain with their respective supply chain. Especially in M2M communications, mobile retailing for industrial automation sector; and next-generation automobile electronics ecosystems development in the automobile ecosystem. Complex regulation landscape in the EU and in Member States may slow adoption of technologies when compared with more lightly regulated jurisdictions. However may also develop higher standards which become a competitive advantage over time. Fragmentation in the market and a weak link between EU research, academia and industry, especially at the graduate level in healthcare sector. Threats Increasing competition from established semiconductor vendors with aggressive investments to retain leadership position. For example, companies such as Freescale and Renesas in the automobile embedded sector. Lack of growth and investment. A risk that the lack of an efficient internal single market and the delays in developing common regulations for all EU member states may disadvantage EU enterprises and reduce their competitiveness, internally and on the global markets. Source: IDC Automotive Sector: The automotive sector is undergoing a phase of fast evolution and innovation, driven by the introduction of a range of new services and applications enabled by car electronics, through the adoption of multicore technologies and intelligent systems. Therefore the revenues of embedded processors and multicore are expected to grow faster in the EU than in other word regions. The EU multicore processor revenues are expected to reach 1.7 billion in 2020, with a 35% CAGR. Key market for Europe transformation in automotive market and introduction of new services create opportunities for European players, existing as well as new ones, to further develop. 17

18 Heterogeneous systems are becoming critical as IVI and telematics demand high level operating systems and automatic control devices and advanced driver assist systems require real-time OS. Development of these stacks is an opportunity for Europe. Similarly next-generation vehicles (such as PHV and EV) are also in the beginning stages of market adoption and represent a market opportunity for EU companies given the established automotive ecosystem of branded vendors and sub system companies and technology suppliers. Potential for new companies or consortia to consolidate development work, important to have standardization and cross industry co-ordination to maintain European strength. The major incumbents in the industry include Renesas, Infineon, Freescale, ST Microelectronics, and TI, which collectively control more than 80% of the market, with about 25% of market revenue coming from EU semi companies (primarily from ST Microelectronics, Infineon, NXP, and Robert Bosch). Industrial Automation Sector: The use of embedded processors and multicore processing will increase dramatically in the industrial system market, which includes retail, industrial automation, test and measurement systems. Increased automation, introduction of consumer type technologies and IP enablement of industrial machinery and machine to machine (M2M) communication will result in increased technological sophistication in the industrial automation market. This opportunity is not only for European system and software companies but also service providers and telecom companies who will be enabling the connectivity and establishing big data and analytic solutions. In the retail submarket, high-level operating systems (HLOS) will be required, while in the industrial automation and test and measurements submarkets, real-time operating systems (RTOS) will still be needed. To accommodate this duality, it will be necessary to develop innovations such as heterogeneous architectures and heterogeneous software stacks. Industrial electronics products require more stringent specifications (compared to consumer products), support for longer product life, longer design cycles, and long-term partnerships. OEMs in these markets look for one-stop solutions and tight relationships with system integrators which build subcomponents for OEMs. The key challenge for the value chain in the industrial markets will be enabling the support of legacy code for the large established OEMs like Siemens and ABB, who are reluctant to support new technologies and software that is not compatible with their legacy investments in code. Enabling legacy code is a key requirement as many of the existing IA market leaders have large volumes of code which need to be replaced or rewritten. Many OEMs lack the skill sets and resources to support hardware board designs, electronic system software development, reference designs, application development, or global field support for hardware and software. These are typically expected from semiconductor providers or other parties. These support services could form a good opportunity for European businesses. Industrial automation is probably one sector experiencing in the decade to come much change and innovation. Computer vision, augmented reality, advanced robotics, AI applications etc will become a reality. This will be the answer to a double trend: the willingness of Europe to re-industrialize, and the way to do it viably as an alternative to low cost labor: little labor, in highly automated and flexible factories. There, multicore and manycore computing platforms will provide the computing power to achieve these changes. Energy Sector: The EU's embedded processor revenues in the energy sector will grow from 281 million in 2011 to 2.2 billion in 2020 (CAGR: 26%). 18

19 Opportunity in the EU regions will be driven by adoption of smart grids and need for increased efficiency in power generation, transmission, and distribution and desire among the population to go "green" in energy creation. This will drive the adoption of embedded processors and multicore processing in energy generation, transmission, distribution, and monitoring. In the energy market, as in the industrial systems market, processors for monitoring and communicating energy use (at home, office buildings, and factories) as well as billing the use based on new monetizing methods such as time of use tariffs, will drive high-level operating systems (HLOS). New equipment that is needed to increase energy use control, energy efficiency, and energy regulators (such as intelligent electronics devices and remote terminal units (RTUs) will still require real-time operating systems (RTOS). This duality will drive the demand for innovations such as heterogeneous architectures and heterogeneous software stacks. Electronics products for energy markets require more stringent specifications (compared to consumer products), longer product life, longer design cycles, and long-term partnerships. OEMs in these markets look to one-stop solutions and tight relationship with system integrators (which build subcomponents for OEMs), and assignments are based on the SI's local presence, project capabilities, and delivery performance. From the semiconductor and software side, direct relationship with OEMs is required for new product development and new design introduction. Currently there is fragmentation in standards and support for these technologies, greater standardization would reduce the cost of implementing these technologies while promoting an EU wide marketplace. More importantly, the business models behind the various "energy transitions" that Europe is experiencing at various paces are not yet clear. Open innovation, and discussion along the value chain, and with institutions in charge of regulation, will be necessary to take advantage of today's strong European position in energy. As systems become more sophisticated they become heterogeneous systems and skills in designing and programming such systems need to be developed. Healthcare: This end-user market includes medical applications of semiconductors and circuits used in portable and fixed medical devices. The EU's embedded processor revenues in the healthcare sector will grow from 38 million in 2011 to 140 million in 2020 (CAGR: 15%). Correspondingly, the EU multicore opportunity will reach 29 million in 2020 (CAGR: 25%). This is actually a tiny portion of the overall healthcare technologies market value: as for automotive, embedded systems drive the performance and innovation of a much larger market of highly complex and competitive machinery and systems. A number of factors are driving demand for technology enabled healthcare. Ageing populations, combined with the high cost of traditional healthcare, will drive increased use of telecare systems providing remote availability of healthcare services (such as home patient monitoring) and greater efficiencies in areas such as patient records management. The EU industry may find the opportunity to grow by taking advantage of the demand in (a) affordable healthcare, enabled by technology innovation (b) increased use of telecare systems, (c) advances in image processing/video processing technologies moving to medical equipment regime and (d) advanced IT systems for patient records keeping, billing and insuring their availability globally and wirelessly. Increased digitization and improvements in image and video processing enable more accurate diagnosis at lower cost. The healthcare market is however one of the most fragmented and not standardized, while it has considerable socioeconomic importance its relatively small size makes it less of an opportunity than other markets. 19

20 Aerospace Industry: Introduction of Integrated Modular Avionics in transport aircraft (since A380) has shown an increase in efficiency and in management of critical applications through key elements for incremental certification (from technical point of view, tools and processes). A prime objective for Airbus for future avionics, and a subject of active research, is to move towards all-integrated Modular Architecture (IMA) avionics, as a means to provide more processing power per unit of volume, weight, and electrical power. An important consequence of the emergence of IMA in aerospace architectures is to allow development of fly-by-wire distributed applications, composed of a great deal of equipment (sensors, actuators, physical devices, software modules, memory modules, etc.) supported by the IMA platform. That leads to more and more complex embedded systems. The increase in safety constraints and the complexity of aircraft/ground systems interactions at same time imply to consider security as a key element for any future avionics systems. The introduction of a new-shared computation resource in business jets: IMA-2G addressed by programs such as SCARLETT should even contribute more to business aviation. IMA-2G brings more powerful and efficient CPU (introducing multicore processing units) and larger bandwidth for backbone network. The use of these new capabilities may allow integrating every aircraft functions in this single IMA-2G system: time critical, high performance or open world functions. Advanced collision avoidance systems and stereo navigations systems in absence of GPS when flying low among obstacles can be addressed using optical sensors delivering measurements relative to the surrounding environment. This would require designing piloting or guidance laws compatible with non-metric low-level visual measurements, to inferring 3D information or GPS-like measurements by computer vision and scene understanding. Table 3 Consumer Electronics SWOT Analysis (I) Strengths Leading processor and GPU IP organizations based in Europe (ARM and Imagination, for example) provide focal points for processor ecosystem development. STB innovation, integration, and middleware experience (including Amino, Ericsson, Oregon Networks, Pace, Siemens, and Technicolor) can be combined with knowledge of local regional markets for a strong global competitive base. Large network for academic research that can address the multiple issues facing the processor ecosystem. Strong experience and assets in glass/lcd panel assembly in Poland. Opportunities Growth in the connection between mobile devices and the digital living room will lead to opportunities on many fronts, including: broadcast platforms, connectivity standards, content management, DRM, interface design, and service provision. All of these areas present opportunities throughout the ecosystem, including software and hardware. Growth in 4K video will lead to a new generation of designs for media processors, infrastructure development, and connectivity technology. Also, demand will strengthen for DTVs and STBs that can take advantage for this generation leap in video resolution. Other devices will be 1 2 year followers in supporting this opportunity, but they will address this demand by More pay TV services will support IP STB or hybrid platforms as these service providers enable greater access to Internet content. Opportunities also exist, especially for software firms, to provide new solutions for many processing areas, including: security, UI design, and image processing. Source: IDC,

21 Table 4 Consumer Electronics SWOT Analysis (II) Weaknesses Threats Outside of processing IP, lack of leadership in tablet and e-reader processors and application processing in general. Also, declining resources in home entertainment networking. Also, no regional leadership in operating systems for key application areas. Cultural attitudes contribute to a lack of applied research to speed the production of new technologies. Structural issues, the regulatory environment, and the diverse set of local markets with different infrastructure issues across the EU result in a challenging business environment that slows the development and adoption of new technologies. An erosion of diversified manufacturing in the EU reduces the region's influence in technology and product selection and reduces fundamental experience and knowledge. Emerging processor suppliers from China are trying to establish their position in the market, threatening companies from other regions, including Europe. These firms have lower operating costs, greater access to the growing market in China, and a larger potential talent pool over the course of the decade. These firms also offer greater flexibility to address customer requirements and can leverage resources to drive faster times to market. The macroeconomic situation will be a tremendous obstacle for the EU to address. Uncertainty will drive down consumption and further weaken regional technology and product development. The EU must turn the corner by 2014; otherwise the region's technology providers may not be able to continue to keep pace with their competition. Manufacturing capacity for advanced process technology for processor fabrication is based largely outside the region. Time and distance increase operating costs for processor design firms. For companies with fab capacity in the region, the lack of scale will make it difficult for these firms to keep the EU competitive in this area. Source: IDC, 2012 Consumer Electronics: In Europe, the consumer electronics end product market was 96.1 billion in 2011, up 12% from 2010; with the perspective to grow with a CAGR of 6% through 2020 (this includes the Digital Living Room, DTV, Mobile Devices, STB, Tablet and e-readers and other consumer devices). In the near term, the economic crisis may challenge this growth rate, which is expected to pick up again in the medium term. Media tablets are driving growth, globally and in Europe; without this new category the industry would be contracting. In addition to the macroeconomic situation, other factors hindering growth include the consolidation of single function devices into smartphones; trailing adoption of e-readers; and slower or shrinking growth in DTV revenues due to saturation, oversupply, and lack of compelling new features and content models. Key opportunities for investment within Europe in the consumer electronics industry are STBs, tablets and e-readers, DTVs, and digital living room devices. One of Europe's strongest assets in this industry is the leadership in processing IP which can be leveraged further to enhance the ecosystem supporting consumer devices. Table 5 Communication and Wireless Infrastructure SWOT Analysis Strengths Leading market share in telecommunications and wireless infrastructure equipment industries (Ericsson, Nokia-Siemens and Alcatel-Lucent) with expertise in multicore systems, silicon, and software. Leadership position in low power multicore architecture with ARM which can be extended to communications and networking infrastructure markets. Educational and research institutions capable of delivering leading edge fundamental advancements for multicore for communications infrastructure. Opportunities Fundamental architectural shifts in datacenter and enterprise networks. Wireless Infrastructure upgrades to 4G/LTE followed by LTE-Advanced. Mounting security threats in wireless infrastructure and enterprise networks require application awareness and deep packet inspection. 21

22 Weaknesses Low market share in enterprise and datacenter networking. Multicore expertise in semiconductor and software for communications applications resides internally within larger companies that move at relatively slower pace. Weak investment in semiconductor process technology that is needed for next generation multicore systems means that EU is dependent on external expertise and capability to manufacture multicore semiconductor devices. Threats Economic downturn in EU and other parts of the world forces carriers/service providers to delay infrastructure upgrades. China competitors, specifically Huawei and ZTE, are gaining market share in current EU core strength areas of telecommunications and wireless infrastructure equipment, and potentially displacing the incumbents. EU could miss the current wave of transition to next generation multicore technologies in datacenter and enterprise networking. Source: IDC 2012 Communication and Wireless Infrastructure: This sector includes the main end-user mobile devices and wireless network equipment. The availability of powerful smartphones and tablets is creating a mobile device revolution. Devices will continue to be the main driver of change in this industry, which in turn drive the evolution of the underlying networks both in the enterprise and data centre environment as well as service provider's telecommunications and wireless infrastructure. Particularly in the telecommunications switching and routing equipment and enterprise networks segments, in security appliances and optical transport equipment. The LTE (long-term evolution) segment of the wireless infrastructure market is expected to increase demand for multicore outpacing all other segments. This is the standard supporting high-speed wireless data for mobile and end-user devices. Opportunities are in machine to machine (M2M), location-based services, premium content and value-added video and voice services, which will increasingly rely upon the underlying networks based on multicore systems, multicore semiconductors and specialized software to take advantage of the available hardware resources. With a low market share of the EU industry in the enterprise and datacenter networking industry, there is a threat that the EU industry may miss on the current wave of transition to next generation multicore technologies Embedded Systems Industrial Sector Summary This chapter analyzed the main end-user markets of multicore technologies, on the basis of their main trends, drivers of demand of multicore, market forecasts from 2011 to 2020 (worldwide and for the EU). After analyzing each market in detail, it is now possible to compare the key findings. While each market has its own specificities, there are interesting commonalities emerging from the research. First of all, the selection of the end-user markets was based on their relevance as the most important segments for the worldwide semiconductor market (as shown by Figure 4) but also on their potential of growth for multicore technologies. We can conclude that all these end-user markets present a positive dynamics of demand for multicore technologies. Figure 5 summarizes the size of each industry in 2011 (the horizontal axis), expected growth to 2020 compared to 2011 (the vertical axis) and the resulting size of the industry in 2020 (the size of the bubble). 22

23 Figure 5 Comparative Analysis of Multicore End User Industries Growth Potential Note: Growth: absolute growth compared to 2011 Source: IDC, 2012 As this figure shows, the consumer electronics market is the largest market however more growth is forecast in sectors such as automotive, energy and industrial automation. Figure 6 below shows the growth from 2011 to 2020 (horizontal axis); the relative strength of Europe in each industry (vertical axis) and the size of the industry in 2020 (relative size of bubble). The European relative strength was calculated through responses to the survey underpinning this research and the analyst teams' work on each industry. Figure 6 Analysis of Strength and Growth of Multicore End User Industries 23

24 Strength: a qualitative assessment of the competitive positioning of the EU industry in each of the markets examined, based on their market dominance, growth dynamics, breadth of offering portfolio, market leadership, capability of innovation and research. Source: IDC, 2012 This chart shows that the opportunity for Europe splits into four groups: Automotive and energy are areas of medium size but strong growth and attractiveness. Industrial automation and healthcare are somehow smaller but are attractive. Communications is larger but Europe has significant strength, however it is slower growing. Consumer is large but slow growing and Europe does not have significant strength. For the segments where the EU already has strengths and have higher growth rates we note the following: The Automotive industry, not surprisingly, presents the largest potential opportunity. Notice that the market is divided in two main subsegments, roughly of the same size: in-vehicle infotainment systems (IVI) with passenger and commercial telematics systems (PCTS) on the one hand, and automatic control devices (ACD)/advanced driver assist systems (ADAS) on the other hand. In both subsegments multicore is expected to become the dominant technology by However, in the first subsegment new competitors from the consumer mobile markets are expected to enter the market, while in the second case the specialized vendors already dominating the industry may be more likely to benefit of the transition to multicore. In addition, the car telematics systems require high-level operating systems, while control devices will still require real-time operating systems. This duality will require the development of innovations such as heterogeneous architectures and heterogeneous software stacks. Mixed criticalities systems will play an important role. As we will see, this is a common element to most of these markets, together with the security challenges. The energy market is next to the automotive one in terms of size of the potential multicore revenues. The growth potential is driven by the adoption of Smart Grids and the need to introduce more sophisticated management and control systems in energy creation, transmission and distribution, also to the final user (smart meters). In this segment too there will be a duality between HLOS and RTOS, requiring the development of heterogeneous architectures. The incumbents of this sector, according to IDC, should not fear too much from external competitors, given the high level of specialization requested and the close relationship with the clients. The industrial market, which includes industrial automation systems, retail systems, test and measurement systems, will see 100% penetration of multicore technologies by 2020, but will generate lower revenues than the other market segments. This market is in proportion a smaller market, highly specialized by subsegment. The retail systems demand (POS and kiosks) is particularly sensitive to the overall economic climate and to the acceptance by consumers. Here too there are issues of heterogeneous architectures and governance of complex systems. Finally, in the healthcare market, demand for multicore will be driven by new medical equipment and by the potential diffusion of telecare systems. However the estimated market is minuscule. The demand for multicore is expected to grow slowly, because of the higher barriers typical of this highly regulated market. It should be noticed that in this market, as in the automotive ACD/ADAS segment, a sizable portion of the market will remain in the 32-bit multicore MCU domain. In sum, there are two large market segments with moderate potential growth, but very high revenues, which are consumer electronics and communication and wireless. And there are three other smaller market segments, automotive, energy and industrial automation, where multicore technologies are only starting to penetrate, and where the opportunities for the 24

25 EU industry are very interesting. It is not surprising that these three segments fall within the broad area of embedded systems, where the EU has a traditional expertise. A further consideration is that these segments are not fixed and in future the boundaries between them will overlap. What is becoming important is less the industrial domain and more the intelligence underpinning it. Smart devices and systems are often no more used for a single industrial domain only but are shared to implement functions and services from different domains. Examples are the smart phone that is currently used for entertainment and communication which can be used for medical applications if properly qualified, or the car which becomes an electronic platform for many services beyond the original task of individual transport. There should be a rising awareness of this from these players, and of the importance of having closely related technology providers General-Purpose and High-Performance Computing The primary focus of this report is on embedded and industrial markets; however a number of trends which affect the general purpose and high performance computing markets will affect embedded markets. The key trends noted in this research are: The majority of hardware, software and datacenter facility innovations come from the Web, at an accelerating pace. Amazon, Microsoft, Facebook, Google and others run datacenters hosting tens of thousands and, in some cases, hundreds of thousands of server nodes, and their demand has been ramping up pushed by the skyrocketing amount of data and applications delivered to consumers and to businesses through the Internet. Being huge consumers of hardware, very large cloud service providers (CSPs) see their datacenter infrastructure as their competitive advantage, as well as one of their main items of capex and opex. For those companies, profitability and ability to succeed in the market are tightly bound to the way they operate their IT backend, which has led them to be the focus and, in many cases, the propeller for most of the innovation in enterprise datacenters over the last ten years, from the use of containerized facilities to implementation of open-source software to analyze large quantities of unstructured data on the fly (e.g., Hadoop), to the adoption of Flash-based storage to the accelerated processing of online queries. Software IP will be increasingly monetized through delivery of services and hardware. Large industry players in the consumer space (Apple, Google) have posed serious challenges to the traditional (Microsoft-like) model of monetizing client-side software through license fees. Good software is now taken for granted and monetized through hardware (Apple) or services (Google) sales. The same is happening in the back-end, with companies like Amazon, Red Hat, Salesforce and others that develop strong, industry leading software, but monetize it through subscribed services, be those support (Red Hat) or consumption (Amazon, Salesforce)-based. IDC believes traditional software players (Microsoft, Oracle, and SAP) to be slowly adapting to this by looking into SaaS delivery models (e.g. SAP Business ByDesign), hardware sales (Oracle Exadata, Microsoft Surface tablet) or capacity sales (e.g. Microsoft Azure IaaS, PaaS). Software innovation happens in or through the public cloud. Large cloud service providers (Amazon, Microsoft, Google, and Facebook) and startups, both B2C and B2B - able to create a business with just an idea and use of scalable cloud resources - will drive most of the software innovation - and at an ever increasing pace. Incumbent will try to acquire to stay the course, but will at times lag trends and be sidelined. Cloud changes customer relationships towards computing hardware. In a traditional onsite infrastructure deployment, the buyer-supplier relationships are such that the end user is a customer of both infrastructure hardware suppliers, as well as of the eventual systems integrator and software supplier. In the public cloud delivery model of SaaS and PaaS types, the end user is effectively only facing one supplier of platform, software or integration services, who in turn is the one handling all low level 25

26 infrastructure and hardware issues. This means that back-end computing hardware and low-level software layers remains increasingly abstracted to the user, gradually losing relevance in the IT strategy of large and small companies. Large cloud service providers absorbing a growing share of hardware spending. While server installed bases on-site grow slowly or remain stagnant, large Web 2.0, online gaming and Infrastructure-as-a-Service providers ramp up their hardware purchases in the region at frantic pace, looking for scalable, dense, and energyoptimized hardware, often deployed in "shots" of hundreds or thousands of units. IDC estimates that already approximately 7% of the x86 shipments in Western Europe in 2011 were headed to cloud service providers, and it expects that portion to grow fast over the next few years. The cost of compute per teraflop keeps dropping. While server shipments to enterprise are slow-growing, and server installed bases rarely expand in floor space, shipments of server x86 CPU cores continued ramping up, and grew 25% year on year in Western Europe in 2011 (versus 3% decline in server units). Cost per core in x86 servers went from more than $1,000 at the beginning of 2008 to around $500 at the end of ARM-based computing will pick up in technical, parallel data processing. Experimental ARM-based server products from HP and Dell show efficient, super-dense and cheaper computing blocks on ARM are coming, but they will take 2-3 years before a consolidated open source software ecosystem rises to support them. Impact of server virtualization on enterprise server demand. Virtualization software is closing in to be deployed on 25% of the new server shipments in mature countries (20% in emerging countries) at the beginning of This is driving demand for more powerful, datacenter-ready server products (typically beefed up 2-Socket machines), but it is also curbing demand for additional server hardware capacity on premise in SMB and enterprise customers in mature countries. There is considerable potential for low power CPUs to enable the creation of more dense datacentres and to reduce power consumption in complex embedded systems (see also the comparison between CPU costs in Figure 11 in this report). Cutting-edge workloads are pulling data back closer to the CPU. In environments where latency is key, including some Web environments, financial high frequency trading and enterprise database or business analytics, there is a trend to bring some, or in some cases (Web specifically) all of the data in internal or DAS storage supports to increase performance. This has spurred storage-centric designs in the enterpriseoriented workload-optimized systems released over the last 24 months (e.g., Oracle Exadata Database Machine, SAP HANA appliances, all accelerated with flash-based solutions), as well as for high-capacity rack servers dedicated to low-bandwidth delivery of Web content (e.g., travel Web sites). This is an area where faster internal storage options such as PCIe flash or SSD can find room to grow. Storage and compute blocks are headed for convergence. IDC believes push for performance, large industry players (Intel above all), growing software automation capabilities are pushing the market towards a serverization of storage, i.e. a merger of storage and compute blocks (typically managed separately Key junction is). the usage of storage controllers based on x86 CPU technology, which will bottom up absorb more and more of the dedicated software capabilities in proprietary storage platforms, thus commoditizing those. Penetration of low cost open source storage management tools (e.g. Red Hat, Nexenta etc.) competes with proprietary storage software at the top, and allows white-box JBODS (just bunch of disk) storage, as well as storage-as-a-service in cloud to become competitive alternatives. Networking hardware will enter x86-fication and commoditization phase within three years. Proprietary network hardware platforms are a cost liability for large Web providers, who need enormous volumes to operate. They have started using low-cost white-box gear combined with networking virtualization software (e.g. Ncira) and IDC 26

27 expects this trend to gradually extend to enterprise datacenters over the next decade, dragging network gears into an accelerated commoditization cycle (potentially combined with use of x86 chips). Storage growth continues unabated. In 2011, the amount of information created and replicated surpassed 1.8 zettabytes (1.8 trillion gigabytes) growing by a factor of nine in just five years. While 75% of the information in the digital universe is generated by individuals, enterprises have some liability for 80% of information in the digital universe at some point in its digital life. Storage requires space, servers require dense power. Storage volume growth is major reason for expansion of datacenter floor space now, whereas server miniaturization requires unheard of datacenter cooling density. Density optimized servers (multi-node rack chassis) are the new standard for extreme compute/watt performance, overcoming blades, and cover almost 5% of the server shipments in Europe. E.g. the latest SeaMicro products pack more than 1,000 computing cores in a 42U rack (256 cores in 10U chassis), for a (very efficient) power draw of 14kW per rack. Standard enterprise datacenters serve between 1kW and 4kW per rack in average. Within HPC we observe the following trends: Many-multicore cards become standard in HPC. GPGPU cards are used to add compute power in blades, density optimized and custom HPC environments. Latest Intel Xeon Phi announcement of x86-based many-cores will further accelerate this as less recompiling is needed. A single card produces 1 Tflop of computing power; competing GPGPU products (e.g. NVIDIA TeslaTM S1070 Computing System) can already be purchased in bundles of many teraflops. Modern HPC hardware with large numbers of CPU cores, each with decreasing levels of memory and memory bandwidth, is causing a mismatch with existing application software, driving a need to fundamentally redesign and rewrite HPC application software for greater parallelism, in order to perform well on future HPC systems. As a result of the primary focus on highly parallel hardware systems, today only about 1% of HPC application codes can exploit 10,000 or more processor cores. The largest HPC hardware systems today contain more than 1 million cores, and much larger supercomputers will begin to arrive before the end of this decade. Without fundamental software innovation their power cannot be harnessed General-Purpose Server Market The server market is a key component of the general computing market for business. According to IDC, in the EMEA region (Europe, Middle East and Africa) the server customer revenue decreased by 1% year on year from 10.5 billion in 2010 to 10.3 billion in This was driven by solid growth performance in the x86 market, up 2% annually, offset by negative growth in the legacy non-x86 business of UNIX and mainframe platforms. The EU 27 market accounted for 8 billion of server customer revenue in 2011, down 4% year on year, and corresponding to approximately 73% of the EMEA market (Figure 7). Volume servers priced at < $25,000 continued to hold the overwhelming majority stake in the market, with 64% of total revenue, or 7.1 billion. In the forecast period to 2016 virtualization and multicore processor technologies will enable customers to migrate higher-end enterprise workloads from Unix servers and mainframes to x86 server platforms. These factors will affect retirement rates and will continue to accelerate the retirement rate of high-end, non-x86 systems. IDC expects a modest decline in the EU27 server market estimating a 5-year CAGR of around -0.8% for combined x86 and non-x86 revenue in Based on the expected commoditization curve, server shipments in the EU27 should remain roughly flat in 27

28 the forecast period (CAGR -0.4) with HPC and cloud service providers replacing mid- to small businesses as consumers of technology (Figure 8). The deteriorating macroeconomic conditions in the Eurozone and in the U.K., combined with high penetration of server technology in mature countries, will continue to exert pressure on demand. Looking at the server revenues by country in 2011, it can be noticed that Germany, France and UK together represent 60% of the revenues in the EU27. This means that the server market is more concentrated in the larger markets than it could be expected based on their relative economic weight in the EU27, as particularly large multinational and IT services companies consolidate their datacenters in fewer, core geographies (Figure 7). Figure 7 Server Market Revenues in the EU27 Compared to Rest of EMEA, EMEA: Europe, Middle East and Africa. Rest of EMEA: Middle East, Africa, Ukraine and Russia. Source: IDC 2013 Figure 8 Server Market Revenues in the EU27, Historical 2011 and Forecast , Millions and Shipments Source: IDC Final Considerations on the SWOT Analysis Analyzing the SWOT tables of the market segments, we find many similarities: in most of these segments the EU has leading user industries, and only a few technology champions. The best known, of course, is ARM and its ecosystem, but there are a few others. 28

29 It should also be noticed that in all these industries, the adoption of multicore is not simply an incremental innovation (improving speed and power): it enables a new level of processes integration and innovation, accompanied by sophisticated applications and services, with advanced control and management capabilities. In all these segments for example the need to develop heterogeneous architectures and software for the coexistence of radically different operative systems emerges clearly. There is also a clear concern about the fragmentation of the European market, and the risk that too burdensome regulation (designed with the best intentions to protect safety, security, privacy...) will prove again a competitive disadvantage for the EU industry and a barrier to innovation. In terms of business model, it is also clear that the multicore innovation will to some extent disrupt some of the "cozy" relationships between suppliers and industry users of microprocessors, focused on dedicated chip designs. In all these sectors, multicore technologies will throw open formerly closed systems, and because of the need to manage vastly more interconnected and interoperable systems, will create new competition perspectives challenging established players. While proximity to the client and in-depth knowledge of vertical markets will remain strong competitive advantages, no EU player will get by without revising in-depth their business model and substantial investments. The need for investments is mentioned particularly in the communication and wireless industry, a traditional stronghold of the EU industry, where there is concern that European players are missing the wave of innovation driven by the enterprise and datacenter networking industry to next generation multicore technologies. The comparative analysis of the SWOT confirms that the window of opportunity opened by the multicore innovation is not limited to a collection of market niches; rather it cuts across the most important sectors of the European economy. It is therefore relevant and useful to identify the horizontal technology and market challenges to be overcome to enable the EU industry to exploit the multicore opportunity. This will be the focus of the next chapter. 29

30 3. M A I N C H A L L E N G E S A N D A R E A S O F A C T I O N 3.1. Overview The phase of development of the microprocessor and multicore solutions market offers an extremely interesting but challenging opportunity to the EU industry and economy. The main trends of the multicore market summarized above show how important this challenge can be: at the very minimum, the EU industries must be able to access and use at best multicore technologies and solutions appropriate for their business, in order to develop their competitiveness. But it would be even better if the EU succeeded in establishing an open and thriving supply ecosystem, able to support next generation system requirements, a flexible leading semiconductor manufacturing network, silicon integration capabilities, and software tools development capabilities. To achieve these goals, we have selected 6 critical research and market development challenges which, in our view, are the most relevant. They are of course interconnected and also to some extent interdependent, but require different types of actions to be solved. Figure 9 below presents a summary view of these main challenges, in a simplified "stack" reflecting the structure of the value chain of the multicore ecosystem. The figure should be read in the following way: The bottom layers illustrates the challenges concerning the design, development and production of multicore processors: they concern the issue of next generation fab capability in Europe and the opportunity to build around the most successful semiconductor ecosystem in Europe, around ARM. These are both technology and market challenges, and are the most important for the competitiveness of the EU multicore industry in the first place. The mid layers focus on the most critical research and technology development problems affecting the adoption of multicore in end-user industries: dealing with heterogeneous hardware and software architectures and the need to invest in the development software tools and systems suited for parallelism. Therefore these challenges affect both the EU industry and the user industries, therefore have a strong potential impact on the innovation capacity of the EU economy. The top layer presents one of the constant European challenges, the need to remove barriers to growth for high-tech SMEs in the multicore market, encouraging the creation and development of start-ups and spin-offs. This challenge is relevant both for the industry and the market development. Finally, there is one critical challenge, possibly the most relevant for the competitiveness of the EU industry and the development of the user market at the same time, which addresses the need to develop vertical integration network models, and (by definition) cuts across the value chain stack presented in the figure. This challenge has truly a systemic impact on the computing ecosystem affected by multicore innovation. 30

31 Figure 9 Main Research and Market Challenges 31 Source: IDC, 2012 More specifically, the challenges can be described as follows: Develop a vertical integration network model. European strength is in relatively small volume industrial markets which use semiconductors. This causes Europe to be a follower in the development of technology which impacts on the ability of Europe to have a strong semiconductor related industry. The transition to multicore is threatening the traditional business models of EU industries in the markets where they are stronger, the embedded market. The fundamental challenge for Europe is achieving the R&D and productive scale able to gain back leadership, or at the very least be competitive in the support of the EU user industry. To achieve this, a possible approach is to radically change the organization of the industry, by developing vertically integrated platforms between specialized suppliers to provide multicore technologies and applications to end user industries. Develop next-generation fab capability in Europe. Maintaining competiveness and carefully managing the intellectual property at each stage of the value chain will be critical factors affecting European actors in the broader multicore-manycore market. A key challenge is determining how important it will be for the EC to invest in nextgeneration fabrication technology and facilities, how important this will be in helping keep IP, skills and expertise in Europe, and the potential impact on product innovation and time to market. There is not an easy answer to this issue, given the scale of the potential investments needed. This is especially true if available investment is utilized to support multiple and diverging national strategies; in which case the necessary critical mass will not be reached. However, there is a need to maintain in Europe at least some production capability to support the design and development capabilities of the players who have chosen the fables model. Build on the low power ecosystem. Low power systems are the key to future success in the embedded systems market. ARM in particular has been successful in mobile devices and increasingly is moving into other embedded markets and into general purpose computing. IDC believes that there is an opportunity to create a Linux-based software ecosystem especially around ARM, both leveraging existing European players (such as Canonical and the SUSE community) and supporting the growth of start-ups that can build applications to run on top of the platform. The challenge will be to develop further this ecosystem producing value added for Europe, building on the competitive advantages developed by this successful company. Support the development of heterogeneous hardware and software architectures. The challenge for Europe will be to harness existing competencies in

32 both hardware and software to develop architectures that are aligned with requirements in key industries such as automotive, healthcare, and energy while avoiding the trap of an over-proliferation of architectures that will dilute resources and impede progression in other areas such as the development of vertically integrated networks. Future multi-manycore architectures will be or include complex generic computing platforms (together with ad hoc OS, middleware and programming tools). This means that the "computing" domain that Europe has fled or has been ousted from will have to be re-entered, using any available opportunities for innovation. This is a challenge for incumbent players; disruptive initiatives must be encouraged as ARM was encouraged in the '90s from European programs. Enable the development of new software tools and systems suited for parallelism. The move to parallelism creates a number of challenges for software development. According to IDC, highly parallel hardware systems cannot be exploited without fundamental redesigning and rewriting application software for greater parallelism. Although multicore processors have been mainstream for several years, current software development still lags behind, with sequential applications still the dominant program design. There is a need for new, cost-effective software tools and systems enabling the development of new software or reengineering of the existing software, as well as investments in new software applications and services. For the EU software industry, this is an important opportunity to gain back lost ground and position itself more strongly for the next decade of innovation. To face this challenge, it is necessary to invest in the research and development of software tools and systems, as well as of the necessary skills and competences, overcoming the fragmentation of the EU scientific and research communities in this field. Remove barriers to growth for high-tech SMEs, supporting the development of start-ups and spin-offs. The market transformation driven by multicore technologies creates dual challenges for European SMEs, on the supply side (to develop innovative multicore technologies and services) and on the end-user side (to exploit innovation based on multicore). The challenge for Europe is to find a way to promote and support a thriving ecosystem of fast growing SMEs some of them becoming major players, in the multicore industry, leveraging strong technology and know-how transfer from research and university to the market, facilitating the emergence of start-ups and spin offs. In the past years, Europe generated a high number of start ups in the semiconductor market, but very few of them have become successful and grown, contrary to what happened in the U.S. and Taiwan. This is a complex market where the needs of embedded systems and the semiconductor industry vary by industrial sector, historic areas of strength and weakness in different parts of Europe, availability of resources such as skills and financing and other factors. These challenges, the possible actions to be taken and their relative impacts are presented in the following paragraphs. 32

33 3.2. Developing a Vertical Integration Network Model As this report shows, the transition to next generation semiconductor technologies will be a disruptive one both for the semiconductor industry and for industrial users of semiconductors Description of the Challenge European industrial users of semiconductors face a number of obstacles to future success, many of which were described by respondents to the survey carried out for this report and in the workshops. The principle issues are: Slowing growth in European markets. The short-term relative weakness in European GDP compared to other world regions may be temporary, however longer term growth is also likely to be slow due to the relative maturity of the region and factors such as demographic pressures. In forecasting long term GDP growth the OECD model suggests 2.4% GDP growth between compared with 3.8% for the world as a whole and 6.6% and 7.6% for China and India respectively. The result for European suppliers is that they do not have the more dynamic home markets that the United States and developing countries can use to test product concepts. Uneven export performance. Coupled with a weakness of growth in European markets it is clear that currently some countries have strong export performance whereas others are weaker. For example comparing 2010 with the pre-recession year of 2007 we find that high tech exports (taken from UN Comtrade database with high tech defined as those with a high R&D intensity such as aerospace, pharmaceuticals, computing, etc) were 4% higher in 2010 than in 2007 for the EU as a whole, 27% higher in France, 3% higher in Germany but 12% lower in the Netherlands and 3% lower in the U.K. For the EU as a whole the export picture is reasonable but member states are moving at different rates with respect to their high-tech exports. Scale of purchasing. European companies are typically smaller than their rivals and hence command less ability to shape markets, including semiconductor markets to meet their sources of competitive advantage. Even in sectors where Europe has some very strong competitors they are typically smaller than the U.S. equivalents. For example, taking the automotive market (data from the OICA World Motor Vehicle Production Statistics), in 2011 European headquartered firms comprise 3 of the top 10 providers by production volume, and have a vehicle production slightly higher than the equivalent U.S. firms, however in the U.S. the 2 firms in the top 10 produce on average 44% more vehicles than their European equivalents. Although the distribution of firms varies by industry those surveyed for this report felt that overall Europe had less scale than other regions. Threat to existing European semiconductor firms. Many industries are using providers of custom designed ASICs for their products. Particularly for the products which are expected to be long in production and support, which are the critical industries for Europe, semiconductor suppliers benefit from long established relationships and relatively high profit margins. As industrial sectors look to adopt multicore it is likely that they will move away from custom designs and use widely available silicon combined with a mix of off the shelf and custom software. There is therefore a threat to existing providers in Europe who currently serve these industries with custom designs Possible Actions 33

34 The fundamental challenge for Europe is achieving the R&D and productive scale which other regions can achieve either through a relatively more consolidated industrial supply chain compared with Europe or through greater state direction of investment. One approach to achieving scale, while respecting the strengths of the European market, is to consider building closer relationships between existing European firms and potential entrants such as start-ups within the same sector who face the same challenge of moving toward multicore. The ultimate goal is to build a common platform for the different sectors in the embedded industry. Depending on industry the platform could take different forms and could evolve over time. The main forms that we expect in ascending order of commonality are: Future requirements definitions. The industry would provide an audit of where the majority of providers are currently engaged in their use of multicore and related technologies and a roadmap to the features and functions that they expect will be needed over the next 10 years. This could be used by the semiconductor industry to develop its product offerings. This is a codification of current practice, as one semiconductor industry representative commented the industry is very reactive to customer needs and does not tend to develop new products to sell to customers. Clearly such a definition would benefit all firms that used it, including those outside of Europe; however by illustrating the requirements of European firms it may help support European firms more than those in other markets. Reference architectures. In this model the platform goes beyond future specifications and looks at how these could be implemented conceptually so that they integrate with the development path the industry is taking. This would provide the semiconductor industry with a better defined roadmap for its products and would allow firms within the semiconductor industry to decide where to specialize. Common standards. This is more prescriptive version of the reference architecture approach. Here the industry players define which areas of the embedded stack are not areas of competitive differentiation for the next 10 years and write standards which they use in their own designs. The semiconductor industry then is able to design their products to take advantage of these standards. Common code. This is a development of the common standards approach. An industry using semiconductors works with the semiconductor industry to write a code base which provides a platform for differentiation. The semiconductor industry then can respond with tools for implementing and optimizing that code and with the areas of differentiation that each player in the end user of semiconductors is looking to focus on. Depending on the needs of each industry we envisage different structures could be used to achieve this goal, these include: Industry working groups. This would be the loosest structure consisting of a committee comprised of the semiconductor using industry and the semiconductor industry. Staff would be loaned from working group members to the working groups. Industry networks. This would consist of a funded organization similar to existing standards bodies where industry participants pay a subscription which pays for a dedicated staff. Incorporated bodies. This would consist of the creation of a company to take forward industry needs with the subscribing organizations as shareholders in that company. The process of implementing this approach is shown in the figure below. 34

35 Figure 10 Approach to the Development of Vertical Integration Models Source: IDC, 2012 Assuming some actions are to be taken, the steps for implementation are: Industry definition. The common definition of industry is too broad for this activity. For example, medical systems include a range of functions such as scanning, analysis, imaging and so forth. It may make sense to consider these as a whole because they benefit from being more integrated or it may make sense to consider each one individually as more focus on each element will result in a stronger offering. Working groups will need to consider the best approach and there needs to be an overall framework to identify commonalities across industries. Agreement on key features. Whatever format the common platform takes the industry will need to have some level of agreement on what is included in that platform. Where embedded systems and multicore are of less value to the end product and hence offer less opportunity for differentiation it is likely that the key features can cover a larger proportion of the total use of semiconductors. Where there are significant opportunities for competitive differentiation it is likely that the features are a smaller set. Agreement on the platform mechanism. There is a need to decide which of the approaches to platform outlined above is suitable for the industry. Clearly each has benefits and disadvantages in terms of the competitive position of individual firms and the opportunities they give to the semiconductor industry. Research and implementation. The industry will need to have a program to research the technical and market challenges for the industry and make decisions on where to focus. As indicated in the chart each structure when implemented will have trade-offs in the scope of the activity, the timeliness of implementation and the usefulness to European industry of that implementation. These factors should be considered in deciding which structure to follow. There are a number of other considerations which should be taken into account in deciding whether to implement this action and in designing the structure: There may not be sufficient opportunities to leverage development across areas in which Europe is currently strong. In particular industries may feel that they do not share sufficient priorities and technical needs with other industries to allow for effective collaboration and may not perceive a need to do so with competitors in the same 35

36 industry. Before putting in place a mechanism to allow this integration there needs to be a process for collecting both technical areas that multiple industries may benefit from collaborating on and views of how to make that collaboration more valuable. Some combination of a horizontal approach to solving future technology challenges and a vertical approach to applying that within an industry might be advantageous. A standards based approach brings the risk that the standards lag behind industry development. Focus needs to be on concrete measures that drive European industry forward rather than trying to codify what is already occurring. Intellectual property protection may inhibit collaboration. This may occur in two ways, firstly through firms not agreeing common terms for collaboration or, secondly, a license being adopted which then limits future use of the technology by business. There is a danger that this approach will lead to a focus on developing technologies and solving problems that are so cutting edge they ignore the benefits of adopting simpler technologies. There is a risk that Europe maintains strength in developing technologies but companies based in other regions are better at adopting them. In developing the approach there needs to be an evaluation of the market benefits of adoption both to existing European companies and start-ups and the capacity of existing industry to absorb these technologies Example Structure There are already moves in some industries to build a common platform. For example in automotive GENIVI is a consortium of companies whose goal is to define a standardized common software platform for developing in-vehicle infotainment (IVI) systems and to nurture a development ecosystem around that platform. IVI systems are system found in high-end cars, commercial vehicles, and increasingly in mid-range cars, that provide a range of entertainment and other functions. The consortium was announced in 2009 with members including automotive manufacturers (BMW Group, PSA Peugeot Citroën, General Motors), tier 1 automotive suppliers (Delphi, Magneti-Marelli, Visteon), Silicon vendors (Intel), and operating system vendors (Wind River Systems, a subsidiary of Intel). Current there exist more than 150 members in GENIVI with ISVs (independent software vendors), middleware vendors, and software services companies with an interest in the automotive sector. GENIVI's goal is to define a standardized common software platform, a packaging of operating system and middleware components that implement a range of nondifferentiating functionalities that all IVI systems require. One of GENIVI's goals is to decrease the cost of developing the base system so that developer resources can be devoted to innovating at higher levels in the software stack, such as the human-machine interface. Therefore Linux and open source software were chosen as software platform and the consortium announced plans to make all of its proprietary code publicly available to the open source community. This is one model for achieving the pooling of resources needed to lower cost and share development work Potential Impacts The research for this report indicates that if no action is taken in this area Europe will fall further behind other regions in both the innovativeness of sectors using semiconductors and in its indigenous semiconductor industry. The reason for this is that other regions have the scale or structures to enable targeted semiconductor investment. If implemented this measure will provide Europe with a structure for investment and, in time, should take 36

37 advantage of the scale of European industry and overcoming the barriers of Europe having smaller firms. Further this measure will reduce design costs for European firms by having common platforms that they can utilize in their products. It will provide a way for new entrants to enter the market as they will not have to build up capability in areas which do not differentiate them in the market. Finally it will provide Europe with the opportunity to lead the next generation of standards and specifications for embedded systems across a range of industries Developing Next-Generation Fab Capability in Europe Description of the Challenge Maintaining competiveness and carefully managing the intellectual property at each stage of the value chain will be critical factors affecting European actors in the broader multicoremanycore market. A key challenge is determining how important it will be for the EC to invest in next-generation fabrication technology and facilities, how important this will be in helping keep IP, skills and expertise in Europe, and the potential impact on product innovation and time to market. Any discussion on the development of next generation fabrication capabilities in Europe will likely focus on the perceived need to invest in capabilities in order to produce components based on the emerging 450mm wafer technology. The industry shift to 450mm wafer technology (from smaller 200mm and 300mm wafers) is a natural evolution that is driven by the need of the semiconductor industry to address economic, productivity and environmental challenges driven by the dictates of Moore's Law. In this respect a discussion on investment in the next generation of fabrication facilities fits well with the a component of the EU 2020 Strategy which states "Europe must promote technologies and production methods that reduce natural resource use, and increase investment in the EU's existing natural resources." However, previous industry transitions to a larger wafer size have brought disruptions and fundamental geo-economic changes in the structure of the industry. New standards are often driven by those actors at the leading edge of the transition and industry concentration will tend to centre on their chosen areas of geographic investment. There is also a belief that the shift to 450mm technology could be the last such shift in the semiconductor industry for a considerable time (perhaps THE last), and may define the geographic makeup of the industry for the foreseeable future. The arguments for significant, long-term investments in next generation (specifically 450mm) fabrication capabilities and the impact on the EU are extensively covered in the final report of SMART 2010/062: "Benefits and Measures to Set Up 450mm Semiconductor Prototyping and to Keep Semiconductor Manufacturing in Europe." One of the conclusions of this report is that "Europe will lose advanced and competitive SC manufacturing infrastructure without a European long-term industrial vision guiding and enabling the coordination of all stakeholders. Such a long-term vision shall not oppose 300mm or 450mm but rather consider both in parallel as part of an advanced manufacturing continuum, taking into account all the stages of the SC supply chain." These sentiments are echoed in the white paper "The Move to the Next Silicon Wafer Size" produced by the European Equipment & Materials 450mm Initiative (EEMI 450), which was presented to Commissioner of the Digital Agenda for Europe, Neelie Kroes, in February, In this paper the EEMI 450 outlines the benefits for the European Equipment & Materials industry of preparing for, and investing in, the transition to 450mm technology (independent of the availability of 450mm fabrication capabilities in Europe). 37

38 In general, full blown investment in next generation fabrication facilities is likely a "bestcase" scenario and part of a broader, long term strategy that impacts the complete European semiconductor industry. However, developments in semiconductor materials science continue, and the introduction of promising new technologies such as Graphene could prove to be serious disruptors for the semiconductor industry. Given the technological uncertainty and the enormous investment required, the discussion around 450mm wafer fabrication is clearly a long-term issue and should not sidetrack the short and medium-term decisions that need to be addressed to maximize the multicore-manycore opportunity. In the context of multicore-manycore technologies, fabrication is one element of the value chain that also includes materials, equipment, process expertise, design tools, software and applications. In the shorter term building an ecosystem that preserves European IP and competiveness in key value-chain segments in the absence of a strategy for nextgeneration fabrication is likely to be a critical step. Over the past two decades, the fastest growing U.S. semiconductor suppliers have adapted to a fabless business model in order to remain at the forefront of process technology and tool development and to support their core competency in design. European actors can look to leverage this proven fabless business model especially in the higher value segments of the semiconductor industry, such as CPU, GPU, FPGAs, advanced SoCs and future multi and manycore architectures. Currently there are a number of European commercial actors (e.g. NXP, ST, Infineon, DSP Group, CEVA) that have the internal intellectual property, system expertise, and foundry engagements to drive the fabless model for Europe. However, despite having an established position in their specific market segments, they face obstacles which would ultimately make driving the fabless business model a huge leap of faith. This is where the support of Public Authority actors and directed community actions to provide a framework for success will be most needed to fully protect European intellectual property and maintain competiveness Possible Actions In the Communication document "Preparing for Our Future: Developing a Common Strategy for Key Enabling Technologies in the EU" (Sept. 2009) the Commission identified micro- and nanoelectronics, including semiconductors as one of a shortlist of the most strategically relevant key enabling technologies (KETs) stating, "Micro and nanoelectronics, including semiconductors, are essential for all goods and services which need intelligent control in sectors as diverse as automotive and transportation, aeronautics and space. Smart industrial control systems permit more efficient management of electricity generation, storage, transport and consumption through intelligent electrical grids and devices." This was further underlines by the Communication document "A European Strategy for Key Enabling Technologies A Bridge to Growth and Jobs," which the Commission released in June 2012, highlighting the belief that "The EU is not capitalizing on its knowledge base" and "the EU's major weakness lies in translating its knowledge base in goods and services." Viewed from this perspective, potential actions for safeguarding intellectual property and maintaining competiveness in multicore-manycore technologies including the development of next generation fabrication facilities in Europe should likely align with general recommendations for developing and deploying KETs. Related to this, the final report of the High-Level Expert Group on Key Enabling Technologies, published in June 2011, provides a structured framework for an approach to KETS to improve overall European competiveness. In their report the HLG called the gap between basic knowledge generation and the commercialization of this knowledge the

39 "valley of death." In short they proposed a three-pillar approach to crossing the "valley of death." The three pillars are: 1. The Technological Research pillar: based on the availability of technological facilities supported by research and technology organizations. Here ideas are turned into competitive technologies. 2. The Product Development pillar: Pilot lines and demonstration capabilities supported by industry consortia. Here working products and prototypes are developed as proof of concept before moving to full scale production. 3. The Competitive Manufacturing pillar: key "anchor" companies support competitive manufacturing capabilities. Here globally competitive production facilities allow products to compete on the world stage. Applying this framework to the specific context of the multicore ecosystem the following potential actions emerge: The initial focus should be on the consolidation and further development of the knowledge elements of the multicore-manycore value chain. In particular, the development of heterogeneous hardware & software architectures, and the leverage of existing strengths in software and application development for key industries is important. This is outlined more clearly in section 3.4 Development of Heterogeneous Hardware/Software Architectures for Specific Application Areas, and includes both the structured investment in academic research as well as leveraging the current position of leading European companies to encourage innovation. In terms of crossing the "valley of death" European vendors and technology suppliers should be encouraged to become first movers in the migration toward heterogeneous architectures and systems across the full value-chain. This is particularly true in key industry segments such as automotive, healthcare, energy and telecommunications where Europe already has a strong presence. Following the "valley of death" concept the next step would be to provide the foundation to produce working prototype and pilot projects to test proof of concept from a user performance perspective. In the context of multicoremanycore this equates to investment in device fabrication pilot lines to produce limited quantities of devices at reasonable cost. The challenges of this step, however, should not be underestimated, and any investment in fabrication capability would likely need to be aligned with the broader strategy around the future of the European semiconductor industry and the industry transition to 450mm wafers as discussed earlier in this section. Pilot lines are expensive, need regular expensive upgrading and even subsidized European facilities would likely come under competitive cost pressure as the potential transition to 450mm leaves older fabrication capacity "looking for business." Areas to concentrate manufacturing will be for products in automotive, aerospace, industrial, and government/health related segments because the solutions must be qualified by stringent design parameters like voltage and long product life cycles (and also where European expertise in system safety and reliability will achieve maximum impact). However the types of facilities that support these market segments are not necessarily leading edge and typically are depreciated facilities that offer healthy margins to suppliers. This means that careful consideration must be given to whether a pilot-line initiative should strive to stay at the leading edge of semiconductor technology or aim to focus on older process capabilities that will still provide a robust foundation for product testing and proving. The development of a fully commercial, globally competitive, fabrication facility in Europe is a longer term issue that is inextricably linked to the future of the complete semiconductor industry in Europe. Seen in the context of the Commission's view of the strategic relevance of micro- and nanoelectronics (incl. semiconductors) KET, and the importance of this KET across a wide variety of industry and social segments, the 39

40 development of a commercial fabrication capability takes on a whole new dimension. However, the huge cost of such an investment and the disruptive nature of new developments in semiconductor materials such as Graphene, means it is very unlikely that Europe could single-handedly take on this challenge. Alternatively, attracting one of the major global players to invest in a European facility (e.g. Samsung or Global Foundries) may make competitive sense. However, purely in the context of multicoremanycore technologies, the development of a large commercial fabrication capability is of lower priority as an action than investing in the more knowledge-led elements of the value-chain. In summary, to sustain the competitiveness of the EU industry particularly for the embedded systems market, the key factor is to enable access to fabrication capabilities such as pilot lines and demonstration capabilities Potential Impacts Developing pilot-line fabrication capabilities is a major investment that would require a coordinated approach from both public administration and commercial actors. Depending on the size and scope of the facilities, implementation might require a review and reallocation of available investment resources to provide sufficient support for capabilities that provide value to the complete community. By providing the ability to quickly prototype and validate multicore-manycore designs and components, European organizations will likely accelerate their ability to move from idea to product. This will likely have an energizing effect across the full software and application value chain, offering actors the possibility to move products into the commercial phase more quickly and stimulating innovation. As these products will often be elements of more complex systems (e.g. automobile, healthcare products) improvements in the multicoremanycore value chain feed growth and competiveness in subsequent value chains. The availability of fabrication capabilities also has a direct positive impact on the materials and equipment industry in Europe. While pilot-line facilities might not in themselves form a significant market for E&M actors, they provide the opportunity to e.g. engage with designers, test processes and materials, develop measurement and tooling capabilities, and remain competitive within the global market. This enables Europe to retain and grow the key skills and knowledge to maintain a strong indigenous semiconductor industry. In addition to direct employment, fabrication facilities for pilot-lines would likely have an associated clustering effect. In effect the fabrication facility would be the catalyst for the development of clusters of associated actors with predominantly highly skilled personnel. While pilot-line fabrication facilities would perhaps not have the same impact as a full-scale commercial facility, a well managed network of pilot facilities would offer an attractive breeding ground for new talent and innovation. The final report of SMART 2010/062 "Benefits and Measures to Set Up 450mm Semiconductor Prototyping and to Keep Semiconductor Manufacturing in Europe" quotes the results from two studies that analyzed the economic impact of the development of semiconductor clusters in Dresden and Crolles: The study (Reverdy 2007) of the Grenoble-Crolles ecosystem from 1993 to 2006 showed that, taking all these effects into account, one job in an advanced semiconductor R&D facility generates at least five additional jobs within the economy. The Dresden study conducted by DIW in 2002 indicated that one job in Dresden's microelectronic operation sites (AMD + Infineon, 6300 employees in total) generated 0.7 jobs in the semiconductor supply chain (4183 employees) and 0.8 induced jobs in the 40

41 rest of the economy due to the redistribution of salaries (4824 employees), leading to a multiplication ratio of 1 to Building on the Low Power Ecosystem Description of the Challenge Low power systems are the key to future success in the embedded systems market. ARM in particular has been successful in mobile devices and increasingly is moving into other embedded markets and into general purpose computing. Both general purpose and embedded systems offer opportunities to develop the European semiconductor industry. Within general purpose computing 2011 saw the development of products that can deliver an even higher performance/watt ratio compared to more "traditional," server processor-based density optimized products. The way to achieve this is cram as many small, power-efficient CPUs as possible and bind them together with highspeed fabric. Although these systems are yet to be tested with real world applications early benchmarks indicate potential for significant power savings. For example the figure below, taken from Caldexa, shows a benchmark comparison for web serving workloads (using the ApacheBench benchmark). Figure 11 Performance per Watt Comparison for Caldexa CPU and Intel CPU Source: IDC 2012 As this figure shows there is considerable potential for low power CPUs to enable the creation of more dense datacentres and to reduce power consumption in complex embedded systems. Two CPU architectures are competing in this market: ARM architectures, based on CPU designs by ARM Holdings PLC and implementations by several companies and start-ups (AMD, Applied Micro, HiSilicon, Calxeda, Cavium Networks, Samsung, Nvidia) that license ARM technology Atom x86 architectures, based on designs and implementations by Intel 41

42 These CPU architectures are then used in system designs developed by vendors such as Dell, HP, Seamicro, and others. We believe that they will initially target the following workloads: Hadoop clusters; servers for content delivery networks (CDN); dedicated hosting; Web frontend processing (HTTP request processing). General purpose commercial applications are unlikely to be a target for these systems over the next few years. Regarding ARM: Calxeda is one of the first movers in the ARM space and has already produced more than one wave of CPU/main-board products targeting server deployment. The product is a chip combining processing unit, network and management software. The company raised $48 million from ARM Holding in 2010 and another $55 million from venture capitalists in October No product is broadly commercialized yet but Calxeda announced commercial products from OEMs by the end of A big hurdle for ARM adoption in the very short term is the lack of a 64-bit instruction set in existing ARM CPUs. ARM Holding added the 64-bit intellectual property part to its designs in the ARMv8 version of its architecture, released in It then released first usable designs based on ARMv8 in October 2012 (Cortex-A53 and Cortex-A57). Major ARM-licensees (AMD, Nvidia etc.) expect to release 64-bit ARM CPUs by bit instruction sets are key to tackle even modest technical computing workloads. Regarding Atom: Intel's Atom faces lower barriers to entry than ARM, as the x86 instruction set is currently the most popular in the industry. Although 32-bit is still prevalent in the Atom product line, 64-bit products are available (and used e.g. by Seamicro). However, current Atom processors do not support ECC memory, which is needed in servers and workstations. Intel has announced availability of special ECC-supporting, 64-bit Atom CPUs by end of 2012 or beginning of 2013 (Codename Centerton). Seamicro is one of the first system movers in the Atom space. The company started off by taking regular Atom 32-bits processors used in netbooks/notebooks and developed a strong fabric and management technology to reduce time lapse in communication between the various codes. It started selling full server solutions consisting of 10U high rack chassis embedding high numbers of mini-motherboards and processors (low-power Xeon and Atom). The densest solution developed achieves 384 dual-core Atom 64-bit processors in a 10U chassis (SM HD). IDC expects ARM/Atom based products to see a strong growth rate along the line of general density-optimized servers, with ramp up in higher single-digit % of server spending only after In the embedded systems market even though there are suppliers with large revenue pools coming from the traditional embedded market such as Renesas, Freescale, Infineon, Microchip Technologies, Texas Instruments, and others, there is no dominant supplier like there is today in PCs or mobile phones. The key reason for the balance among suppliers is that MCU and DSP companies differentiate with proprietary architectures, peripherals and support legacy developer investments in programming code. Traditionally the microcontroller market has been driven primarily by 4 and 8 bit demand from white goods, consumer electronics, industrial, and automotive applications. Many of the market segments are deeply embedded markets with long product cycles on one spectrum, and short product life cycles and very low cost requirements on the other. The automotive market has become the largest driver of revenues for MCUs, with more than one-third of overall revenues coming from this market. As we look forward, MCUs will see a gradual displacement from a growing base of ARM solutions in the 32 bit MCU space. Even though the ARM architecture only accounts for a small part of the overall MCU volumes, the MCU replacement market is one of ARM's fastest 42

43 growing licensing and volume segments. Intelligent systems will accelerate this migration over time as MCUs give way to 32 bit microprocessor architectures because there is more demand for 32 bit support, graphical user interfaces, larger displays, standard tools and peripherals, larger memory requirements, and applications that demand scalable performance. Another threat to high-end MCUs are DSPs especially for low-power, batteryoperated applications, since inherently DSPs can be tailored for lower power than MCUs for a given speed. Applications here include audio devices, biometric systems, and portable medical equipment. Coupled with WiFi, DSPs have a potential to be a low-cost participant in the embedded systems market. There are at least three challenges and opportunities here: A key opportunity for Europe is in the software ecosystem for ARM which will develop through 2013 and Regular, off-the shelf Windows and Linux software and applications cannot run on ARM processors and need recompiling, tuning, and optimizing. IDC believes that the opportunity to create a Linux-based software ecosystem especially around ARM is broad, both leveraging existing European players (such as Canonical and the SUSE community) and supporting the growth of start-ups that can build applications to run on top of the platform. Systems built to address the challenges created by low power designs are being developed, currently primarily by existing vendors who are located in the U.S. or Asia. However this disruption creates opportunities for startups in Europe to build a place in the market Possible Action Addressing these challenges and opportunities requires close partnership between industry, national government, the European Commission, and research bodies. The market is developing quickly and although there is time to build areas of strength it needs reasonably swift action. In particular we suggest the following actions: A goal should be to develop a next generation of general purpose computing systems which utilize ARM technologies. This will provide some key sectors for European development, such as cloud computing and high performance computing in financial services, with technology alternatives to the established providers. As well as providing a useful service to European industry such as product could spur further development of European technology firms both in general purpose and embedded markets. Going further the European Commission should evaluate the feasibility of developing research initiatives around specific use cases for ARM systems and low power computing. These systems will blur the boundaries between embedded systems and general purpose systems. Areas such as automotive, aerospace, financial services and bio-life sciences may be suited to specific research programs to evaluate whether ARM and low power systems are superior to existing technologies. We recommend that Europe invest in establishing one or more centers-of-excellence where European software developers can port and optimize their applications for use on ARM processors, and where users of the applications can receive help and experience in running their codes on ARM-based servers. This could give European users a substantial advantage over their non-european counterparts where the exploitation of ARM-based servers is concerned. 43

44 Potential Impact Low power general purpose and embedded systems represent an opportunity for both the European semiconductor industry and users of semiconductors. In the absence of strong support for the European semiconductor industry it is likely that existing global firms headquartered in the US and Asia will continue to maintain their strong presence in the market. By taking advantage of the disruptive potential of ARM there is an opportunity to establish a leading position in the next technology wave Development of Heterogeneous Hardware/Software Architectures Description of the Challenge Reference architectures for heterogeneous hardware or software environments bring advantages in terms of facilitating the development of reusable tools, portability of code and development of new applications. The challenge for Europe will be to harness existing competencies in both hardware and software to develop architectures that are aligned with requirements in key industries such as automotive, healthcare, and energy while avoiding the trap of an over-proliferation of architectures that will dilute resources and impede progression in other areas such as the development of vertically integrated networks, Much of the development of multicore-manycore and system-on-chip technology has been focused on addressing the associated issues of power consumption and heat dissipation. However, the increased number of components that can be physically placed on a chip means that further improvements in power consumption and energy-efficiency must come from the optimization of individual components for specific tasks or applications. While component optimization has a positive impact on these hardware-centric challenges, it also increases the level of complexity associated with developing software that can effectively leverage the hardware capabilities. Currently the overwhelming majority of (legacy) application code is serial in nature and effectively adapting or rewriting applications for parallel environments is complex and expensive. An increasing fragmentation of hardware architectures based on the need to optimize the power performance of specific components for specific tasks further compounds this problem. There is a real risk that the relative ease of optimizing on-chip components will lead to a proliferation of multicore-manycore architectures (possibly based on the developers own personal component preferences) and further widen the gap between hardware capabilities and available application software. The challenge is to balance the multicore-manycore hardware and software development requirements in a way that allows us to maintain the power consumption/efficiency benefits of optimized hardware components while providing a framework to effectively develop parallelized applications that will meet both current and future needs. This argument is underlined in the publication Computing Systems: Research Challenges Ahead. The HiPEAC Vision 2011/2012 published by the High Performance & Embedded Architecture and Compilation (HiPEAC). The report, released in June 2011, indicates that "... specializing computing devices is the most promising path to dramatically improving power efficiency" but that "unfortunately this trend will only worsen the complexity and cost of developing software for these systems." In some ways it can be argued that reference hardware or software architectures for specific application areas are an important element of addressing this challenge. While this view is certainly valid, it also brings into focus the secondary challenge of choosing areas where reference architecture would make both commercial and technological sense. 44

45 Developing reference architectures for applications/market segments that do not have sufficient scale would be hugely expensive. It is also unlikely to be attractive enough for software developers to make the required changes to their applications. The issue is further complicated by the growing application of multicore-manycore technologies in systems where safety is key consideration. Up to now, much of the development of multicore-manycore hardware and software has been focused on the consumer markets where speed and performance are vital commercial success factors, but where there is no inherent need for real time measurement and analysis (although hardcore gamers might beg to differ). Increasingly, multicore-manycore technology is being utilized in industries such as automotive, rail systems and aerospace, where Europe currently has a relatively strong competitive position. It can be argued that these changes offer a chance for Europe to develop a leadership position in safety-critical multicore-manycore systems, with a high level of industrial applicability. However, safety-critical market is an example of a market segment where the commercial opportunity is unlikely to be large enough to support development of reference architecture. This is outlined in the draft report of the Workshop on Mixed Criticality Systems held on February 12, 2012, which states: "Development of multicores based on a reference architecture dedicated to safety critical applications and more amenable to certification is likely to be prohibitively expensive unless cross domain applications are possible to obtain critical mass. There is thus an opportunity to bring different domains together, aerospace, automotive, rail etc. to develop 'certification friendly' multicore devices. An alternative is to develop methodologies and toolsets that allow existing commercial off-the-shelf devices to be utilized in safety-critical applications. Here there is a need to satisfy safety requirements and provide time predictability. This will require developments in the underlying theoretical analysis of systems and the tools to support this analysis to be cost efficient. Here there is a risk of fragmentation across sectors due a wide spectrum of multicore architectures being developed." Possible Action Addressing the challenges posed by the need to develop heterogeneous hardware and software architectures entails more than just a treatment of the individual symptoms. It is critical that any actions undertaken address the issues in a holistic manner taking into account both the specific hardware and software challenges as well as the overarching issues related to integration of legacy applications, portability of applications and the commercial potential for solutions developed. Of necessity, any EC actions to address the challenges outlined here will over time need to include a multitude of industry and academic actors. Developers for instance, will need tools to manage the complexity of developing new applications or migrating legacy applications to new architecture. Achieving this may need the introduction of completely new approaches and systems for programming developed by specialized academic research establishments. Action: Academic Research to Facilitate Industry Improvements As an example, the complex interactions in multicore-manycore systems demand system level tools that provide a comprehensive view of multicore-manycore behavior so that developers can visualize interactions between the cores, eliminate performance bottlenecks and identify opportunities for parallelism so that each core is utilized effectively. Both symmetric multi processing (SMP) and asymmetric multi processing (AMP) modes are useful in the communications infrastructure depending upon the application. Structured investment in academic research into these advanced software requirements deriving efficiencies from parallelism in the code, software partitioning, running different operating systems simultaneously, and associated application software geared towards data or control plane

46 functions would provide a strong basis to support the practical application requirements of industry actors. In terms of developing new reference architectures, it is likely that existing architectures already provide a robust platform for future-oriented enhancements. A strong focus on commonality and reusability should be leading characteristics of research initiatives. However, specific circumstances, such as systems with strict safety and mission criticality requirements (as outlined earlier) will almost certainly require new architectures. The goal should be to enable these investments to deliver results that are quickly applicable within key industry segments, thus delivering measureable impact in the full industry value chain. Care must be taken to ensure that sufficient attention is paid to projects that are crossfunctional in nature combining elements of semiconductor, software, architecture and tools. In addition, further consideration must be given to the scale of projects for investment. A focus on outcomes that are rapidly applicable in industry will almost certainly dictate smaller projects that address specific industry challenges or "niche" applications. However, this needs to be balanced by the requirement to maximize commonality and reusability across critical industry segments. Action: Leverage Existing Industry Leadership A number of European companies have a strong position in the multicore-manycore space, and action should be taken to further leverage the position of these leading players to encourage innovation, enhance emerging ecosystems and provide a foundation for European research and commercial endeavor to reach a wider audience. The recently (June 2012) launched Heterogeneous System Architecture (HAS) Foundation counts both ARM and Imagination Technologies among its founding members (also included are AMD, Qualcomm, Texas Instruments, Samsung and MediaTek). By its own definition: The HSA (Heterogeneous System Architecture) Foundation is a not-for-profit consortium for SoC IP vendors, OEMs, academia, SoC vendors, OSVs and ISVs whose goal is to make it easy to program for parallel computing. HSA members are building a heterogeneous compute ecosystem, rooted in industry standards, for combining scalar processing on the CPU with parallel processing on the GPU while enabling high bandwidth access to memory and high application performance at low power consumption. HSA defines interfaces for parallel computation utilizing CPU, GPU and other programmable and fixed function devices, and support for a diverse set of high-level programming languages, thereby creating the next foundation in general purpose computing. Groupings such as the HSA will likely be key contributors to the development of the ecosystems required to develop the hardware and software architectures needed to maximize the potential of multicore-manycore technologies. European organizations should be encouraged to follow the example of leaders such as ARM and Imagination Technologies and become engaged with or affiliated to newly emerging groupings. Reciprocally, ARM and Imagination Technologies as European leaders should be encouraged to act as evangelists for building a heterogeneous computing ecosystem and help European companies understand the benefits and opportunities that such an ecosystem can bring. This concept of leveraging leadership can also be applied in industries that are impacted by the development of multicore-manycore technologies for example in telecommunications and media. Here heterogeneous architectures comprising a blend of processor cores will become increasingly important in the communications infrastructure as platforms converge to support a variety of traffic. Increasing video content in mobile networks requires incorporating media transcoding capability into the infrastructure for supporting the distribution of high definition video content. The traffic could be originating from video servers in the cloud to support a variety of mobile devices that have limited processing and bandwidth resources. European telecommunications companies such Ericsson are already leading the efforts for next generation video codec standards such as H

47 As video content is expected to consume an increasing proportion of mobile infrastructure bandwidth European companies should be encouraged to maintain a leading position by investing in the several core technologies that will be required end educated on the broader value-chain opportunities afforded by being aligned with a leading player such as Ericsson. Where Europe has companies who play a leading role, their position and success should be leveraged to provide both inspiration and broader ecosystem opportunities Potential Impact Successfully developing heterogeneous architectures for specific application areas benefits the European Community in a number of ways. Firstly, it will create a new technology value chain across the full breadth of technologies from Silicon IP to the software development role. Success will necessitate the development and production of a range of value-added products, tools and services that will support the joint goals of on-chip component optimization and the simplification of application development for heterogeneous systems. A successful development of these value-chain ecosystems will also make it easier to concentrate available skills (particularly related to developing applications for parallel environments) in areas where their impact can be maximized. As we have pointed out, an uncontrolled proliferation of architectures would likely make the challenge of producing optimized parallel applications more difficult and would inevitably lead to a dilution of the available skills. This lack of critical mass in skills would greatly increase the risk of failure in any single application area and could seriously retard the development of the industry in general. Beyond the purely technological value-chains, European leadership in this field would lead to an accelerated rate of commercial product development based on these technologies. Industries such as automotive, telecommunications & media, energy and healthcare where Europe already has a strong international presence are likely the initial key beneficiaries. These are the industries where the convergence of the forces of massively increased data production/storage/consumption, mobility and improved user experience are most obvious and where the benefits of multicore-manycore systems will be most needed. Successful development of heterogeneous architectures for these application types, potentially with an industry leading focus on safety and reliability, will be a powerful competitive force in these industries and vital for future competitive success Development of new Software Tools, Systems and Applications Suited for Parallelism Description of the Challenge The move to parallelism creates a number of challenges for software development. Although multi-core processors have been mainstream for several years, current software development still lags behind, with sequential applications still the dominant program design. The underlying reason for this is that parallel programming is far more difficult than sequential programming. In particular, the cost and time required to parallelizing existing sequential applications is nearly prohibitive for all but the most performance-critical domains. In addition, it is difficult to manage projects for migration towards parallelism and there is a lack of awareness of the real issues at stake. However, according to IDC, it is no more possible to postpone the problem of software innovation relying only on hardware performance. The gap between hardware and software innovation is becoming too wide. There is a need for new, cost-effective software tools and systems enabling the development of new software or reengineering of the existing software. For the EU software 47

48 industry, this is an important opportunity to gain back lost ground and position itself more strongly for the next decade of innovation. To face this challenge, it is necessary to invest in the research and development of software tools and systems, as well as of the necessary skills and competences, overcoming the fragmentation of the EU scientific and research communities in this field. Until now, the domain of parallel software was restricted to HPC. With the transition toward multicore, the need for parallel software is extending also to the other computing markets. Europe (like the rest of the world) has been overly focused on funding parallel hardware to the detriment of parallel software (as discussed in the study "Financing a Software Infrastructure for Highly Parallelised Codes" produced by IDC for DG INFSO in 2011, based on a thorough analysis of European research centres engaged in the development of parallel codes). Where funding has been made available for parallel software development, the funding typically has been for only a year or two, compared with at least 5-10 years of funding needed to develop robust, production-quality software that can remain useful for or even 30 years and across multiple generations of HPC and technical computing hardware systems. As a result of the primary focus on highly parallel hardware systems, today, according to IDC, only about 1% of HPC application codes can exploit 10,000 or more processor cores, while the largest HPC hardware systems today contain more than 1 million cores, and much larger supercomputers will begin to arrive before the end of this decade. While the main software research challenges are common to all market segments, there will be also a need to take into account their different specificities and to build on the specialized expertise accumulated in time, particularly for the embedded systems market. In addressing the multicore-manycore challenge, it is important that the issue of embedded systems receives a prominent position in the discussion, in particular as it relates to critical systems and safety. As has been previously indicated many of the key industries where multicore-manycore technologies are being increasingly used have robust requirements for measurability and safety e.g. Automotive, Railways, Aeronautics. Therefore in order to effectively support the evolving needs of these key industry segments any investments in software methodologies or tools should also be aligned with the inherent safety requirements of these industries and the need to integrate systems with mixed levels of criticality. Concerning the embedded systems market, a recent study by IDC Design of Future Embedded Systems (SMART 2009/0063), produced by IDC on behalf of DG Information Society and Media (now DG Connect); the final report was published in April 2012) identified two main categories of packaged software and software tools: Software development tools: a macro category encompassing application development software, change and configuration management software, and automated software quality PLM tools: a macro category encompassing PLM software, electronic design automation (EDA) software and content management software Based on current innovation trends, the main research challenges to be faced in the next years about software development tools are: Validation of the correctness and certification of parallel programs Development of parallelization tools "Rapid" development of critical applications (safety) for parallel architectures (multicore/multiprocessor) Development of tools to optimize algorithms for multicore computers Application deployment tools for parallel technologies (multicore/multiprocessor) to minimize validation and certification processes 48

49 The critical research topics related to PLM tools are: Support of critical and non-critical architectures on the same multicore platform Application deployment on parallel technologies (multicore/ multiprocessor) minimizing the validation and certification processes Generic hardware architectures, programmable and massively multicore Clearly, these research topics are complementary and support the research topics identified for the development of heterogeneous architectures and are broadly common to all the industry sectors. Similar challenges and research areas were identified by the FP7 Network of excellence HIPEAC (HiPEAC Vision 2011/2012 published by the High Performance & Embedded Architecture and Compilation Network, by M. Duranton, D. Black-Schaffer, S. Yehia, K. De Bosschere) under three main topics: Move to parallelism and programming models, including aspects such as locality management, optimizations programmer hints, runtime systems and adaptability Development of new compilers, including automatic parallelization, adaptive compilation and intelligent optimization Development of software including virtualization, input, output, storage, and networking, simulation and design automation tools, deterministic performance tools As the HIPEAC roadmap points out, the research on automatic compilers is particularly relevant, since current tools are not sufficiently effective and cannot be applied on a large scale. However, succeeding in their development may have high impacts on the costeffectiveness and the speed of parallelization of existing codes, thereby facilitating the adoption of multicore-based innovation in user sectors Possible Action Define a clear research roadmap to achieve the transition to parallel software, including the timescale of objectives, pursuing both the adaptation of existing software and the research on new programming paradigms. The development of systems assisting with parallelizing existing software and compiler tools is a short-term objective; the development of new programming paradigms and self-adapting software with full performance portability is a long term objective. This research roadmap should be complementary with and contribute to the development of a European parallel software development strategy, able to maximize the returns on the existing and planned hardware investments in the HPC field, such as the PRACE program. The European Exascale Software Initiative (EESI) created a strong, credible framework plan for this development, but the plan will not happen without appropriate funding. Ensure that safety and mixed-criticality are core elements of the research roadmap to ensure that these investments are applicable to the developmental needs of key European industry segments. A leading position in multicore-manycore technologies and embedded systems in critical systems can also provide positive economic impact by allowing European organizations to serve markets and opportunities across the globe. Aim for greater collaboration and networking between the research centers and the fragmented EU industry, in order to achieve a critical mass of research and knowledge. This does not mean simply to leverage cooperative research programs such as H2020, but also to explore the possibility to create a network of centers of excellence for parallel software helping to unify Europe's scientific and engineering research communities around common objectives. The most practical way to do this is by key application domains. Users in the same domain will be interested in the same set of computational challenges and parallel software applications. Each centre of excellence would assume primary responsibility for EC-supported parallel software development in its domain, 49

50 including the creation of plans recommending which parallel codes should be advanced, how and when this would be done, who would lead the effort, and how much funding would be needed. This should also help to bridge the gap between the embedded software and the HPC community looking for common objectives and synergies Potential Impact The opportunities for growth for the EU industry in the software market addressed to multicore are relevant. The value of this market was of approximately 1 billion in 2012 in Europe, projected to grow to 1.8 billion by The PLM tools category is the biggest market and accounts for over three-quarters of the market, while the software development tools category has seen the fastest growth and will expand at a healthy 9.1% CAGR between 2010 and Obviously this is only a niche market compared to the total value of the embedded systems market (estimated at 1 trillion in 2012 worldwide, with potential growth to 1.5 trillion by This includes all external spending by OEMs and subcontractors on technology and services suppliers, including communications equipment and mobile phones). Within the embedded systems market, the opportunity to design new software or reengineer existing software for the emerging applications and SoS is much more relevant. Globalization may open new markets for embedded systems. European ESD players can now enter into BRIC countries, as those markets cannot rely on their own embedded system providers. But if competition from the U.S. is a longstanding threat, it has now intensified with competition from BRIC countries in particular. There is a risk that by , basic know-how in BRIC countries will have reached European and U.S. levels. The EU industry needs to stay a step ahead of the competition. On the other hand, there is a need of an EU-level effort of investment in these research challenges. The EU industry in this area is very fragmented. IDC estimates that there are about 430 European ESD (embedded software design) software vendors but there are no more than 20 companies with turnover over 10 million, which account for nearly the three quarters of the market. Dassault Systèmes and SiemensPLM are the two largest companies. In this context, the EU research programs have an important role to play to achieve a critical mass of investment in research and develop the skills and knowledge needed. Table 6 Worldwide ESD Packaged Software Revenue, ( M) CAGR * Total ESD packaged software revenues 3,311 3,726 4,592 6, % Europe revenues 986 1,091 1,319 1, % CAGR 5% CAGR % CAGR % CAGR North America revenues 1,519 1,702 2,079 2, % CAGR 6% CAGR % CAGR % CAGR Asia/Pacific revenues , % CAGR 7% CAGR 9% CAGR 9% CAGR 15 50

51 CAGR * Rest of the World % CAGR 9% CAGR % CAGR % CAGR Note: CAGR reflects the yearly rate required to go from the starting point to the end point. This metric ignores variation between the end points, as it only uses the end points in the calculation. Source: IDC, Remove Barriers to Growth for High-Tech SMEs Description of the Challenge The market transformation driven by multicore technologies creates dual challenges for European SMEs, on the supply side (to develop innovative multicore technologies and services) and on the end-user side (to exploit innovation based on multicore). In addition, as for every other major technology evolution, there are new opportunities to create start ups and spin offs from research. On the supply side, a few of the fabless companies and specialized chip design companies in Europe are SMEs, usually integrated within complex supply chains. For example, ASML has an integrator business model under which 90% of its systems costs are externally supplied. The company works with approximately 500 suppliers worldwide underpinned by an R&D network of about 20,000 jobs, 35% (by product-related supply value) of which are located in the Netherlands, 45% elsewhere in Europe and the remaining 20% elsewhere. According to a recent study on semiconductor manufacturing (SMART 2010/062: Benefits and Measures to Set Up 450mm Semiconductor Prototyping and to Keep Semiconductor Manufacturing in Europe The role of Public Authorities and Programmes Final report, February 2012) the presence of a European-based 450mm fab offering foundry services would strongly improve the competitiveness of the EU fabless SMEs and would help to develop start-up clusters across Europe. This would be driven by the easier technology access and increased share of foundry mind, augmented by the considerably reduced time and cost of not having to regularly commute between Asia and Europe. However, Europe has a strong ecosystem of SME software providers in the embedded systems field, particularly in the area of safety critical systems. This is mainly due to investment from and collaboration with large organizations such as Airbus and ESA. The SMEs in this area will need to innovate profoundly to keep up with the new challenges for software opened by the transition to multicore/manycore. Another constant problem in Europe is the scarcity of risk funding for start ups and spin offs. Innovative firms, especially in the seed and early-stage when a business is not in a position to generate cash-flows which would allow servicing debt, have almost no other option but to turn to equity investors such as business angels and venture capital firms to obtain the financing they need. But, according to a recent study on SME access to finance, "The nearterm climate for new supplies of angel and VC money investing into Europe's early stage SMEs looks bleak given the backdrop of the current economic crisis. In particular VC fundraising has become increasingly difficult and dropped significantly between 2007 and Institutional investors like banks, insurance companies and pension funds have materially reduced their exposure to this asset class and many existing venture funds and business angels remain focused on supporting their portfolio companies with few meaningful exits being announced." (Report of the chairman of the expert group on the cross-border

52 matching of innovative firms with suitable investors, on behalf of the European Commission, SME Access to Finance Directorate-General for Enterprise and Industry, 2012). Within this context, the challenge for Europe is to find a way to promote and support a thriving ecosystem of SMEs in the multicore industry, leveraging strong technology and know-how transfer from research and university to the market, facilitating the emergence of start-ups and spin offs. In the past years, Europe generated a high number of start ups in the semiconductor market, but very few of them have become successful and grown, contrary to what happened in the U.S. and Taiwan Possible Action The barriers to growth of high tech SMEs in Europe are well known and are related with framework conditions such as a fragmented internal market, insufficient access to finance and skilled human resources, difficulty to achieve a critical mass of revenues, often lack of awareness of innovation potential. There is huge number of reports and policies addressing these problems. Here we will suggest some options for action, which are finalized to the multicore market as described in this report. They are the following. Leverage the presence of strong ICT components industry clusters in Europe to provide support to start ups or small high tech companies through their infrastructures. Economic literature shows that SMEs active in clusters make more swift progress, particularly when partnering larger industrial partners. The clusters are more likely to be examples of the "triple helix" approach (interaction between business, research and university) and to support technology transfer. Providing funding for research, incubators, labs to test new ideas, addressed to SMEs or new businesses opening in industry clusters, should be particularly effective. There is definitely a case to encourage new enterprises development within those clusters with the support of public money to sustain their technology transfer and collaboration activities. Encourage the development of smart coaching networks, learning communities, and innovative forms of "crowdfunding" for seed funding new ideas and help to train the would-be entrepreneurs in the industry. Improve access of high tech SMEs to advanced technologies, such as cutting edge silicon technologies or advanced HPC. This is being experimented in FP7 FoF (Factories of the Future) initiative. Suggested scenarios include ISVs porting their applications to a cloud of HPC resources and experimenting with them in a controlled environment; or supporting the development of a prototype sustainable European commercial cloud of HPC resources in manufacturing and engineering, including the necessary orchestration and access services. Simulation services are the new frontier of industrial research and have never been affordable to SMEs; in the past they could be provided only by highly expensive HPC centers, in the near future the implementation of the cloud model for access to distributed HPC resources will make it feasible for innovative SMEs. The experimentation of sustainable business models and HPC clouds to serve the high tech SMEs market may well be one of the most productive implementations of multicore technologies and parallel software Potential Impact In terms of potential benefits, the creation of a thriving ecosystem both of large and small businesses, enabling the emergence of new start ups and spin offs, would improve the competitiveness of the EU industry. On the other hand, if no action is taken, the current population of SMEs active in the multicore market is likely to continue shrinking and losing ground in comparison with out Asiatic rivals. 52

53 4. T H E S O C I O E C O N O M I C R E L E V A N C E O F M U L T I C O R E I N N O V A T I O N 4.1. Overview The increasing diffusion of multicore technologies is profoundly changing the semiconductor industry and affecting the dynamics of all the end-user markets of multiprocessors and SoC (system on a chip). According to IDC estimates, these technologies will take over the semiconductor market by In all the user industries, the adoption of multicore technologies is not simply an incremental innovation (improving speed and power): it enables a new level of process integration and innovation, accompanied by sophisticated applications and services, with advanced control and management capabilities. Therefore a successful migration to multicore has implications not only for the competitiveness and strength of the industry, but for the overall development of the European socio-economic system. Conversely, the inability to deal with the research and technology challenges outlined in the previous chapters bears potential risks for the EU economy, in terms of potential jobs loss and/or lack of employment growth, and of missing innovation in a wide range of user sectors. The assessment of the potential economic impacts of multicore is difficult without an appropriate macroeconomic model, but based on our research we will focus on the following aspects: The relevance of the multicore ecosystem for the European economy seen through the estimate of its employment "footprint," that is the total number of jobs in the industries involved by the multicore value chain. The concentration of ICT components and systems manufacturing in a small number of key industry clusters in a few EU countries. The potential contribution of multicore innovation to emerging social needs and the major social challenges faced by Europe, as identified in Horizon The Employment Footprint of the Multicore Ecosystem in Europe The multicore value chain extends from semiconductor manufacturing to the embedded systems markets (see Figure 2). By adding up the number of employees in all the industries comprised by this ecosystem we develop an estimate of the employment "footprint" of the multicore ecosystem. Lacking statistical data on specific multicore-related activities, this approach allows estimating the total number of jobs which are likely to be affected by the more or less successful evolution of the multicore market. The employment "footprint" of the multicore industry in Europe covered 2.9 million jobs in 2010 (Figure 12), plus 8 million jobs in the multicore user industries (Figure 13), for a total of approximately million jobs (source: Eurostat 2010). This corresponds roughly to 20% of the total employment in industry sectors in Europe, split between 5% for the supplier industries and 15% in the user industries. This shows that the multicore ecosystem has a high relevance for the European economy and labor market. More specifically, the estimate of the direct employment in the multicore industry includes the following subsectors as defined by Eurostat: Manufacture of electronic components and boards (306,000 jobs in 2010) Computer programming, consultancy and related activities (2.6 million jobs in 2010) 53

54 The first subsector identifies roughly the semiconductor manufacturing industry in Europe, and is tiny compared to the second one. Computer programming and services employs 2.6 million people: obviously only a minority are involved in parallel programming and other software development activities linked with multicore. However, we can argue that all the computer programming and services sector will be impacted by the profound changes in software programming and services driven by pervasive multicore-based innovation. Therefore we believe that we can include these jobs among the "employment footprint" of the multicore industry. Concerning employment by country (Table 7), it is important noticing that Germany and France together represent the lion's share of electronic components manufacturing employment, followed by Italy, while the U.K. has a lower presence there. This is confirmed by the comparison presented in Table 7 between the main countries' share of total employment in the electronics manufacturing sector and their share of the EU GDP (as a parameter benchmark). The table shows that both Germany and Poland have a higher share of electronics manufacturing than their EU GDP share, showing a relative higher specialization in the sector, while France and Italy have a share equal to their GDP share. The U.K. and Spain instead have a lower level of electronics manufacturing jobs than their GDP share would imply. On the other hand, the U.K. is the leading country in the EU for computer services, which is the reason why the U.K. has a large share of multicore direct employment, as shown in Figure 12. Table 7 Employment in Electronic Component Manufacturing by MS, 2010 Country GDP 2010 MS Share of EU GDP 2010 % Employment in Electronic Component Manufacturing MS Share of Employment % EU (27) 11,950, % 306, % Germany 2,462,100 21% 73,570 24% Spain 1,047,103 9% 8,445 3% France 1,917,191 16% 50,466 16% Italy 1,547,117 13% 40,222 13% Poland 322,661 3% 19,963 7% United Kingdom 1,571,205 13% 22,814 7% Rest of Europe 3,082,864 26% 91,106 30% Source: Eurostat

55 Figure 12 Multicore Industry Employment Footprint in the EU, 2010 Total Direct Employment per country (%) EU = 2,913,497 (2010) 29% 19% Germany United Kingdom 4% 7% 9% 13% 19% France Italy Spain Poland Rest of Europe Source: IDC elaborations on Eurostat 2010 The classification by Eurostat does not correspond exactly to the taxonomy used by IDC for the identification of the main multicore end-user sectors; however the following Table 8 shows which Eurostat sectors were included for the estimate presented in Figure 12, mainly consisting of the manufacturing industry sectors. Table 8 Employment in Electronic Component Manufacturing by MS, 2010 Eurostat Sectors Manufacture of computer, electronic and optical products, minus components (includes electromedical equipment) Multicore End-User Sectors (IDC Taxonomy) Computing, consumer electronics, medical, wired and wireless communication (partially) Manufacture of electrical equipment (including batteries, wiring and wiring equipment) Wired and wireless communication (partially) Manufacture of industrial machinery Manufacture of motor vehicles, trailers and semitrailers Industrial and energy Automotive Manufacture of other transport equipment Aerospace and defense Source: IDC

56 Figure 13 Multicore User Sectors Employment in the EU, 2010 Total Employment per country - User sectors (%) EU = 8,036,065 (2010) 27% 33% Germany 5% 6% 12% 9% 8% United Kingdom France Italy Poland Spain Rest of Europe Source: IDC elaborations on Eurostat As shown by Figure 13, and unsurprisingly, Germany has the largest share of the multicore user sectors employment footprint (33%), followed by Italy with 12% (due to Italy's strength in industrial machinery and electrical equipment) and France with 9%. The U.K. confirms its vocation as a service economy, with only 8% of the total multicore user industries footprint. This analysis shows that the multicore employment footprint is a very relevant component of European manufacturing employment. Moreover, this does not consider the indirect effect on the services industries employing these products (from transportation services, to healthcare, to the utilities). Overall, a successful introduction of multicore innovation, leading to healthy growth in the industries of the ecosystem, has a relevant overall potential impact on employment growth in Europe The Role of ICT Component Manufacturing Clusters The concentration of the multicore industry in Germany and France is confirmed by an ongoing IDC study for the EC (SMART 2011/0063 Strategies for innovative and effective ICT Components & Systems Manufacturing in Europe, Draft Final Report, IDC, November 30, 2012), which identified and interviewed 48 industry clusters specialized in ICT components and systems manufacturing. As shown by Figure 14 below, 17 clusters are in Germany, 6 in France, 5 in the U.K. (according to the Community Framework for State Aid for Research and Development and Innovation "Clusters can be defined as a group of firms, related economic actors, and institutions that are located near each other and have reached a sufficient scale to develop specialized expertise, services, resources, suppliers and skills.") These clusters cover the main phases of the value chain, including IC -MEMS & Sensor Design; Front-End production, process and equipment; Back-End packaging and equipment; Assembly & Test; Industrial products and systems solutions. There are 24 clusters specialized in Front-end production, process and equipment, which includes the implementation of multicore innovation. While most of the clusters tend to be specialized in no more than 2 phases of the manufacturing value chain, there are only 7 clusters covering the entire chain. This means that most players are trying to compete on subsegments, rather than taking on the larger and richest markets. Germany and France host most of the clusters falling in this category. Many of the clusters are relatively recent: 22 emerged between 1999 and 2005 and 21 clusters after As a comparison, most of the worldwide-class clusters analyzed outside

57 Europe (e.g. U.S., Taiwan, and China) have been created in the '80s. The most recent clusters tend to focus on the new areas of research, such as Organic Electronics, Printed Intelligence, Nanotechnologies for ICT components. Figure 14 Mapping of ICT Components Manufacturing Clusters in the EU Source: IDC study Strategies for Innovative and Effective ICT Components & Systems Manufacturing in Europe, 2012 The European clusters are characterized by a strong presence of SMEs, and are less likely to be driven by large enterprises than U.S. or Asian clusters. They compete for and achieve a high amount of public funding, particularly for research and development. From the point of view of this study, there are some interesting initiatives and policies supporting cooperative research and networking which deserve mentioning and could be a model for EU initiatives: The strategic collaboration between the German Silicon Saxony cluster at Dresden and the French Minalogic cluster at Grenoble constitutes a best practice case In Europe for cross-border cluster collaboration. The two clusters have strengthened their cooperation in the area of nanoelectronics and nanotechnologies focusing on education, research and development, industrial deployment, SME coordination, and environment. Their collaboration has now grown into a new project, Silicon Europe, which joins Silicon Saxony and Minalogic with Point-One, centered on Eindhoven (the Netherlands), and "DSP Valley," centered on Leuven (Belgium). The partner clusters in total have nearly 800 members (297 in Silicon Saxony, 204 in Minalogic, 170 in Point-One, 75 in DSP Valley), more than 75% of which are SMEs. They count for more than 150,000 jobs. This is a very interesting example of collaboration which may create a critical mass of research investments and innovation developments. The Spitzencluster initiative in Germany is funding 15 cutting-edge clusters in Germany with 40 million Euros each, to develop collaborative and cutting-edge research projects. The federal contribution must be matched by business and private investors. Among the clusters receiving funding through are Forum Organic Electronics which aims to make Germany the world's leading research, development and production location in the field of organic electronics, and Cool Silicon which aims to make communications more climate-friendly and to become one of the world's leading locations for energy efficiency in electronics. The Microelectronics Support Centre at the Science and Technology Facilities Council's (STFC's) Rutherford Appleton Laboratory near Oxford in the U.K. is recognized as an 57

58 established centre of excellence with a 25 year history in supporting the microelectronics industry. The Centre consists of a group of experienced microelectronic specialists who support research and educational activities at more than 600 academic institutions across Europe, for instance in the form of technical consultancy and train-the-trainer courses. In Denmark, the Copenhagen International Business Information Technology Hub (CIBIT) has established an accelerator program with a total budget of 13.3 million, cofunded by the ERDF, with the aim to select 150 would-be entrepreneurs, train them and provide them with some seed funding. The ultimate target of CIBIT Accelerace Invest is to create at least 100 high-growth companies in the region and increase regional employment in the ICT sector with 5000 people by the end of So far 50 of the entrepreneurs participating in the program have managed to raise pre-seed capital for their companies. These clusters are active either in the domain of multicore technologies or in complementary domains, and their presence confirms once again the potential capability for growth and innovation of the EU industry in these markets. Conversely, supporting research and innovation in the multicore area may contribute to the strengthening of these clusters, which are drivers of economic growth and employment creation in several key European regions Meeting Social Challenges Multicore-driven innovation is bound to generate a wide variety of new products and services in all the end-user markets. Many of these will respond to current or emerging social needs, by improving individual or public welfare. The following tables summarize briefly the main new technologies and systems based on multicore and SoC innovation, which will respond to main social needs. The following considerations build also on the research carried out by IDC in the 2012 report on "Design of Future Embedded Systems." (SMART2009/0063, Automotive The emergence of next-generation vehicles represents a disruptive innovation in the automotive sector and a critical step in the roadmap foreseen by automobile manufacturers. Embedded systems will underpin automotive SoS performance in sectors such as electric vehicles (EVs), transportation as a service, and Car 2.0. EV Fleet trials in the EU (there are around 16 of them) will gather real-world use-based information (e.g., EV efficiencies, charging profiles). New business models will be tested, and key drivers and technological challenges are summarized below. 58

59 Societal Needs Reduce carbon footprint/improve sustainability Improve mobility Meet consumer expectations of service quality and life style continuity Greater choice and personalization through product/service differentiation Improved safety Easier driving for elderly and disabled citizens Automotive Systems/Technologies Smart charging, vehicle-to-grid (V2G) incar technology (software, communications) Charging infrastructure Wide area networks/gps Automated driving Back-office systems billing and settlement systems CRM and account management systems Analytics (demand forecasting, GHG emissions) Enterprise asset management Driver/consumer service portals Mobility Social media Healthcare In the healthcare sector, increasing demand due to ageing and prevalence of chronic diseases drives higher costs and is contrasted by the need for reducing public budgets. Embedded systems for healthcare cover a wide range of equipment supporting treatment and recovery of patients, and encompass the patients themselves and medical professionals supporting them. As a result, these systems need to operate as part of a dynamic network of interoperable systems and therefore should be user friendly and highly reliable. Societal Needs Satisfy the healthcare demand of the ageing population Manage increasing healthcare costs without reducing quality and care Deal with the prevalence of chronic disease Meet patients expectations of service quality and life style continuity Manage provider staffing shortages that are significant and accelerating Healthcare Systems/Technologies Telemedicine and remote patient monitoring Intelligent/connected medical devices (glucometers, pulse oximeters, blood pressure monitors) Wide/home area networks Care management systems (to enable remote care by clinicians) Medical equipment and e-health Systems that support early diagnosis and prevention: high performance, embedded computers are having a huge impact in medical imaging Electronic medical record (EMR)/personal health record (PHR) systems Patient portals Telepresence/videoconferencing Energy The energy market is next to the automotive one in terms of size of the potential multicore revenues. The growth potential is driven by the adoption of smart grids and the need to introduce more sophisticated management and control systems in energy creation, transmission and distribution, also to the final user (smart meters). This will support the 59

60 introduction of smart grids enabling the integration of renewable and distributed energy resources, reducing the carbon footprint of the industry and improving customer service. Societal Needs and Drivers Reduce carbon footprint/improve sustainability Reduce energy consumption and prices Need to enable energy efficiency and demand response programs Integration of renewable and distributed energy resources Energy Systems/Technologies Smart Grids Smart meters, grid sensors Neighborhood/wide area networks (wireless, mesh, WiMax) Meter data management and grid management systems Multicore processors and SoC will also contribute to the development and implementation of smart building and home automation technologies. Societal Needs and Drivers Reduce carbon footprint/improve sustainability Reduce energy consumption and prices Need to enable energy efficiency and demand response programs Smart Building Home Automation Systems/Technologies Energy efficiency and demand response In-home displays, smart thermostats, smart appliances Home area networks Customer relationship management (CRM) systems, analytics, and customer portals Industrial Automation and Retail The industrial systems market includes industrial automation systems, retail systems (kiosks and POS, points of sale), test and measurement systems. Similarly to automotive, industrial systems are increasingly moving from discrete, isolated real-time processing using RTOS (real-time operating systems) to connected high-level processing with advanced microprocessors and HLOS (high-level operating systems). Key drivers for this trend are increased industrial automation, increased efficiency requirements of factories and plants, cost reduction roadmaps of various industrial OEMs, and a migration toward advanced retail kiosks. This will in turn improve safety, improve work conditions and productivity, and satisfy demand for better and more personalized service. 60

61 Societal Needs and Drivers Improve work conditions and productivity Improve safety Improve product quality and personalization Meet new consumers needs Industrial Automation Systems/Technologies Connected high-level processing with advanced microprocessors and HLOS (high level operating systems) Introduction of consumer-driven user interfaces such as touch control, advanced graphics, and voice controls into human-machine interfaces Growth in machine-to-machine (M2M) communications Retail systems market (kiosks and POS) Customer relationship management (CRM) systems, analytics, and customer portals Remote management of POS Summary View of Social Challenges The following table summarizes the potential contribution of multicore innovation to emerging social needs and the major social challenges faced by Europe, as identified in Horizon Table 9 Multicore Innovation meeting Social Challenges in H2020 Potential Impacts of Multicore Innovation by Sector Health, Demographic Change and Wellbeing Secure, Clean and Efficient Energy Smart, Green, and Integrated Transport Inclusive, Innovative, and Secure Societies Climate Action, Resource Efficiency, and Raw Materials Automotive XX XXX XX Healthcare XXX XX Energy XXX XX XXX Industrial automation and retail X X XX Legend: X = low impact; XX = medium impact XXX = high impact Source: IDC,

62 5. C O N C L U S I O N S A N D R E C O M M E N D A T I O N S The main conclusions and recommendations presented here have been discussed and validated at the final workshop on November 28, Overview In the previous chapters we have carried out a thorough assessment of the changes to the hardware and software market that are driven by the current use, and predicted growth, of multicore and heterogeneous system on chip processors. We have estimated the potential growth of multicore/manycore processors and shown how, in the period up to 2020, the diffusion of these technologies, together with the adoption of systems on a chip (SoC) and heterogeneous systems, will change profoundly the computing ecosystem. This will affect the competitiveness of the European semiconductor industry and of the end-user markets such as the embedded systems market. We have then examined in detail the main challenges faced by the EU industry and economy due to this wave of innovation. This chapter builds on the analysis of the main challenges and potential actions to be taken to solve them and presents IDC's recommendations for European Commission intervention, addressed both to the EU semiconductor industry, involved in bringing multicore systems to the market, and the main users, with specific attention paid to the embedded systems market articulated in its main segments Recommendations There are two main common themes characterizing the main challenges identified in the previous chapter: The first is the need to overcome the fragmentation of the EU industry and market, linking together its pockets of strengths, focusing investments and efforts in selected priority areas where there is a chance to make a real difference for European competitiveness, rallying around the few successful companies. At the same time, the analysis has shown that all the challenges identified are not specific to single market segments or products; rather they tend to be horizontal challenges with broad impacts. Technology innovation is raising the bar for the industry; this does not mean that market niches or specific competences are going to disappear, but that some challenges cannot be won playing alone or counting on the occasional maverick company. Our recommendations therefore suggest building on collaborative and coordinated efforts, looking for economies of scale and scope. Horizon 2020 does seem an appropriate instrument to respond to these challenges. The second theme, but by no means less important, is that the absolute level of investment should be raised. Even before the economic crisis, the European industry has underinvested for years and has been retreating from the most demanding segments of the semiconductor ecosystem (particularly manufacturing). However, with the widespread diffusion of multicore/manycore, it is clear that without considerable investments in next generation hardware and software systems the EU semiconductor and embedded systems industry risks to decline into irrelevance, exposing the EU user sectors to the risks of dependence from global suppliers without any local alternative. On the other hand, the good news is that the potential for success exists, if Europe will accept the challenge to build on its existing strengths and competences. The main recommendations are the following. 62

63 Business Model Innovation and Reorganization of the Value Chain The European Commission should promote the development of vertical integration in the semiconductor industry, by promoting the creation of industry working groups and collaborative networks with the goal to develop common requirements, common reference architectures, common standards, and common code base. This collaborative approach would not eliminate competition; rather the players would compete starting from a common platform and a common basis. A possible example is the GENIVI consortium of companies whose goal is to define a standardized common software platform for developing in-vehicle infotainment (IVI) systems and to nurture a development ecosystem around that platform. This would allow directing a critical mass of R&D investments to these common platforms. Research and Innovation Concerning Horizon 2020, the EC should take full advantage of the new program's extended focus on research and innovation, based on the conceptual model of the "three pillars bridge" cutting across the valley of death to bring innovation from the lab to the market, to promote the development of multicore technologies. Enabling and industrial technologies will be one of the priorities in H2020; the scope of the ICT themes on "A new generation of components and systems" and "Engineering of advanced and smart embedded components and systems" will include the development and implementation of multicore/manycore technologies. There is however a need to include in these themes a strong focus on software and applications developments, in order to enable the actual diffusion of these technologies as discussed above. The recommendations are structured in two main groups: recommendations about the scope of research to be conducted in H2020 and about the way research should be organized. Research Scope More specifically, the EC should focus H2020 research and development activities relevant for multicore/manycore on the following main themes: Development of cost-effective, advanced software tools and systems suited for parallelism and new applications, enabling the development of new software or reengineering legacy codes. To do so the following research areas should be prioritized: o Short term: Systems to enable understanding of existing code and assist with parallelizing. Directive-based compiler tools. Runtime analysis. o Medium term: Programming systems with a more integrated runtime and language environment. Hardware virtualization. Providing object-oriented paradigms across accelerators. Integrated performance/power/timing modeling and optimization. Performance and service guarantees for safety critical and mixed critical requirements. o Medium-long term: Develop new methods to approach natural data processing, as required also by the diffusion of Big Data. o Long term: Full performance portability. Self-adapting software. New programming paradigms. Research into the development of heterogeneous hardware and software architectures to harness existing competencies in both hardware and software to develop architectures that are aligned with requirements in key industries such as automotive, healthcare, and energy while avoiding the trap of an over-proliferation of architectures that will dilute resources and impede progression in other areas such as the development of vertically integrated networks. 63

64 This type of research should address the development of crosscomponent/cross-layer optimization for design integration. This requires for example system-level design decision making, cross-layer static and dynamic optimization, a focus on programming systems rather than devices and on "virtualizing" the hardware(s). The need to rethink the global architecture of the systems, in light of energy, technology constraints, and modern workloads, will require research into innovative methods to organize the memory system hierarchy, insure protection and security, reliability, availability, and serviceability, and develop algorithms for unreliable hardware. Today's ISA (instruction set architectures) are insufficient to express the requirements of today's heterogeneous multicore platforms Languages can therefore not express the semantic intent. Research should address the development of new ISA that can define the new hardware for many existing languages. Research into advanced software requirements deriving efficiencies from parallelism in the code, software partitioning, running different operating systems simultaneously, and associated application software geared towards data or control plane functions would provide a strong basis to support the practical application requirements of industry actors. Ensure that safety and mixed-criticality are core elements of the research roadmap to ensure that these investments are applicable to the developmental needs of key European industry segments. A leading position in multicore-manycore technologies and embedded systems in critical systems can also provide positive economic impact by allowing European organizations to serve markets and opportunities across the globe. Organization of Research Promote a system of research centers of excellence (rather than networks) differentiated by specialization (or domain) in order to focus on the most relevant EU competences. Promote research and development projects focused on the development of precommercial new applications and services based on multicore innovation, which could address the vertical markets targeted by embedded systems. This line of research could complement and counterbalance the initiatives on consolidating research centers by scope and developing vertical integration platforms, by leaving room to support cutting-edge applied research in the main vertical sectors typical off the embedded systems market. In terms of developing new reference architectures, it is likely that existing architectures already provide a robust platform for future-oriented enhancements. A strong focus on commonality and reusability should be leading characteristics of research initiatives. However, specific circumstances, such as systems with strict safety and mission criticality requirements (as outlined earlier) will almost certainly require new architectures. In terms of crossing the "valley of death" European vendors and technology suppliers should be encouraged to become first movers in the migration toward heterogeneous architectures and systems across the full value-chain. This is particularly true in key industry segments such as automotive, healthcare, energy and telecommunications where Europe already has a strong presence. Define a clear research roadmap to achieve the transition to parallel software, including the timescale of objectives, pursuing both the adaptation of existing software and the research on new programming paradigms. This research roadmap should be complementary with and contribute to the development of a European parallel software development strategy, able to maximize the returns on the existing and planned hardware investments in the HPC field, such as the PRACE program. The European Exascale Software Initiative (EESI) created a 64

65 strong, credible framework plan for this development, but the plan will not happen without appropriate funding. Developing Pilot Lines and Research in Cutting-Edge Manufacturing Technologies in Europe There seems to be little chance that Europe will invest in full-scale fab capability for next generation 450mm microprocessors manufacturing. However, actions should be taken to maintain competitiveness of the EU industry and carefully manage the intellectual property at each stage of the multicore/manycore value chain, sustaining the competitive positioning of fabless operators. There is also research to be done on potentially disruptive technologies and developments in this field. To do so, the EC should: Increase the investments in device fabrication pilot lines and prototypes to produce limited quantities of devices at reasonable cost, to be accessible for high-tech SMEs and any other European player interested in the design and development of new multicore/manycore. Give careful consideration as to whether a pilot-line initiative should strive to stay at the leading edge of semiconductor technology or aim to focus on older process capabilities that will still provide a robust foundation for product testing and proving. This choice will be heavily influenced by the requirements of fabless actors and their target markets. Developing an inventory of requirements should be considered part of the overall development roadmap and should be based on an understanding of the existing and future commercial opportunities for European fables actors. Areas to concentrate manufacturing will be for products in automotive, aerospace, industrial, and government/health related segments because the solutions must be qualified by stringent design parameters like voltage and long product life cycles (and also where European expertise in system safety and reliability will achieve maximum impact). Align strategic decisions with the broader strategy around the future of the European semiconductor industry and the potential industry transition to 450mm wafers. Carry out research on how potentially disruptive new technologies/materials such as Graphene will change the semiconductor landscape and impact the European industry. This is potentially a separate research program which is relevant to the multicore-manycore topic specifically, but has implications for the full semiconductor industry in Europe within the context of horizon 2020 and beyond. Building on the Low Power Ecosystem There is one leading EU player in the low power technologies field ARM which has been able to develop a worldwide business ecosystem. The EC should look for ways to build on this ecosystem in order to multiply the potential competitive advantages for the whole EU industry and expand the value added for Europe. To do so the EC should, within Horizon 2020: Carry out research on the governance needed and the challenges created by low power designs, which are being developed in US and Asia, primarily by existing vendors located in the same world regions. However this disruption creates opportunities for startups in Europe to build a place in the market. Support the recompilation of software needed to run in the software ecosystem for ARM, which will develop in the next few years. Regular, off-the-shelf Windows and Linux software and applications cannot run on ARM processors and need recompiling, tuning and optimizing. Promote the development of a Linux-based software ecosystem revolving around ARM technologies and support the creation and growth of start-ups that can build applications to run on top of the platform. Remove Barriers to Growth for SMEs 65

66 To implement this recommendation we suggest the two following strands of action. Promote the development of start-ups and spin-offs in advanced multicore technologies In order to create the conditions for start-ups and spin-offs to be created and successfully grow in Europe the EC should: Leverage the presence of strong ICT components industry clusters in Europe to provide support to start ups or small high tech companies through their infrastructures. This could include providing ad-hoc incentives, making sure that IPR policies are favorable to commercial exploitation of research, providing incentives for the establishment of incubators, organize and education and training programs. This may require to support the updating of the rules for state aid to R&D&I. For example, the High Level Expert Group Report on Key Enabling Technologies (KET HLG, 2011) recommended the introduction of a matching clause into general EU State Aid rules, which would allow Member States to match funding up to the maximum levels of support provided elsewhere for product development and manufacturing activities while respecting WTO rules. This may be required to enable funding in this area. Encourage the development of smart coaching networks and, learning communities, to help to train the would-be entrepreneurs in the industry. Make sure that start-ups and spin-offs have access to the latest advanced silicon technologies at affordable prices, including IDM production lines. Build the bridge between research and market opportunities The H2020 program strategy is to help close the gap between research and market innovation in the field of multi-and manycore technologies. To do so the EC should: Develop awareness campaigns on multicore benefits and promote the development of business and use cases of adoption of innovative services based on multi-manycore. This may be focused on advanced innovation in specific end-user sectors (such as smart cars, smart homes, smart cities...) or be focused on the adoption of manycore technologies by high tech SMEs. Continue the initiatives started in FP Work Programme through the Factory of the Future PPP, to develop and test simulation services for engineering/manufacturing SMEs over HPC-Clouds. This will enable SMEs to access cutting-edge industrial research tools and support EU competitiveness in industrial manufacturing. Support and enable the pre-commercial procurement of prototype systems, to bring research closer to the problems of exploitation. This should also help start-ups or hightech SMEs to get reference customers and gain a commercial foothold in the market. Promote education and skills in parallel programming and multicore technologies Given the lack of skills particularly in parallel programming mentioned by most stakeholders, the EC should: In the field of R&D&I policies, contribute to the development of awareness raising and communication initiatives promoting microelectronics education and careers as being attractive, prestigious and a future "job for life" with both personal and financial reward. It is also important to increase the awareness and attract students to a wide range of technological disciplines, as microelectronics should not be considered in isolation in the context of education. The EC should cooperate with industry and national government to increase industry based training and certification (IBTC) offers for up-skilling ICT practitioners but also for retraining side-entries, in order to respond to the emerging demand for parallel programming and other eskills necessary to the multicore market and industry. 66

67 Start an initiative for the development of European guidelines and quality labels for new curricula in the parallel programming domain, building on the activities of the ICT Skills Workshop of the European Standardization Committee (CEN), in particular the development of the European e-competence Framework (e-cf) together with the activities underway in the Educational Programme of the European CIO Association with its Executive MBA and Business Architecture courses and using the terminology, definitions, business or IT approach compliant with the CEN e-cf and the CEN professional job profiles. 67

68 M E T H O D O L O G Y A N N E X Semiconductor market research IDC's research methodology for the microprocessor market encompasses qualitative and quantitative methods to study the market from a top-down and bottom-up perspective. The top-down perspective emphasizes data gathering from major microprocessor IP companies, major MPU and SoC suppliers, and major embedded software vendors. IP vendors include ARM, MIPS, Imagination Technologies, and Tensilica. Embedded MPU and SoC suppliers include Freescale, Intel, IBM, Applied Micro, Broadcom, Marvell, Renesas, TI, Qualcomm, Sigma Design, Sony, Core Logic, NXP, Toshiba, Fujitsu, Cavium Networks, NetLogic, Tilera, LSI, Samsung, Mediatek, MStar, Oracle, ST Microelectronics, Infineon, and PMC-Sierra. Embedded Software vendors include Microsoft, Wind River, Greenhills, Montevista, QNX, and Enea. The top-down research includes surveys that gather trends, insights on markets, pricing, road maps, and actual shipments/asps/revenue. Data is gathered to determine revenue and unit volumes shipped in Interviews were conducted to build assumptions to support forecasts through The bottom-up approach ties existing system forecasts to MPU usage ratios and penetration of architectures. Using device ASVs enables checks on MPU and SoC ASVs and overall BOM costs and spending on systems. MPU and SoC data was then further segmented by processor architecture x86, Power Architecture, MIPS, ARM, SH, and others and by component type standalone processors, ASSPs (SoCs), and ASICs (SoCs). Microprocessor and systems data was then further analyzed to size each major systems market and establish single and multicore market size estimates throughout our forecast period. Stakeholders Interviews and Workshop In order to collect the stakeholders' opinions, IDC organized a set of telephone interviews and a workshop "Multicore Technologies Trends 2020 and Opportunities for Europe," which was held in Brussels on May 31, During May 2012 to August 2012, IDC conducted interviews with industry experts and academics to discuss the main challenges and opportunities impacting the hardware and software market driven by the current use, and predicted growth, of multicore and heterogeneous system on chip designs. The interviews ran for approximately 30 minutes and were based on a questionnaire designed by IDC. The interview guides for industry and research/academia are provided in the following section of this report. In total 29 interviews have been completed and we are working to complete an additional 4 interviews on segments or companies which we believe should be represented. A further 8 companies were approached for interview and they declined believing that they couldn't contribute to the research, 4 companies were interviewed and are not included in this research as the information they provided did not meaningfully add to the research base. The table below provides the list of participants who have been interviewed. 68

69 Table 10 List of Companies Interviewed Company Country Sector Airbus France Aerospace ARM U.K. Embedded/semiconductor BMW Car IT GmbH Germany Automotive Broadcom U.S. and Asia/Pacific Mobile wireless components/telecom Cassidian Germany Aerospace Cavium U.S. and Asia/Pacific Communications/telecom Continental France Automotive Delcam U.K. Industrial Emteco Embedded Technology Communications/SYSGO partnership Germany Industrial Entelios Germany Energy EPCC Edinburgh Parallel Computing Centre U.K. Academia France Telecom/Orange Labs France Telecommunications Freescale U.S. and Asia/Pacific Embedded/semiconductor Gemalto France Communications/telecom HLRS/University of Stuttgart, HLRS High Performing Computer Centre Stuttgart Germany Academia Imagination U.S. and Asia/Pacific Embedded/semiconductor Intel U.S. and Asia/Pacific Embedded/semiconductor Karlsruhe Institute of Technology Germany Academia Liebherr-Aerospace Lindenberg GmbH Germany Aerospace MIPS Technologies Inc. U.S. and Asia/Pacific Embedded/semiconductor MPSA (Peugeot Citroen) France Automotive Nvidia U.S. and Asia/Pacific Embedded/semiconductor Rockwell Collins France Aerospace Samsung Asia Consumer electronics Schneider Electrics France Embedded/semiconductor Shadow Robot Company U.K. Robotics Siemens AG Germany Mobile wireless components/telecom 69

70 ST Ericsson France Mobile wireless components/telecom SYSGO AG Germany Industrial Thales Research and Technology France Aerospace Vienna University of Technology Austria Academia Source: IDC, 2012 Glossary of Terms and Definitions Advanced driver assist systems (ADAS): Automotive systems designed to assist drivers to operate their vehicle more safely. This Includes parking assistance, lane departure warning, adaptive cruise control, and blind spot detection systems. Application-specific integrated circuit (ASIC): An ASIC is a custom semiconductor device sold to a single OEM. Most, if not all, of the intellectual property comes from the OEM in this design. Varieties of custom ASICs include cell-based ASICs, gate array ASICs, new structured ASICs, and FPGAs. Application-specific standard product (ASSP): An ASSP is an off-the-shelf semiconductor device sold to multiple OEMs. Most, if not all, of the intellectual property comes from the semiconductor supplier in this implementation. Automatic control devices (ACD): Electronics used to control, monitor, and manage engine, cooling, entry, windows, security, climate control, braking, airbags, suspension, and other systems tied into the primary functioning of the vehicle. Bill of materials of system (BOM). Compound annual growth rate (CAGR): A measure for capturing annual growth rate over several years, often expressed in percentage. Control area network (CAN): A wired network, primarily used in Automobile and Industrial markets to connect various subsystems for data exchange. Digital signal processor (DSP): A DSP is a specialized processor that performs mathematical operations to manipulate analogue information that has been converted into a digital form e.g. in VoIP applications, DSPs are used to perform voice channel processing, echo cancellation and compression/decompression functions. DSPs are primarily used in analogue systems to process real time data. A-D and D-A converters needed to provide such functionality may sometimes be integrated on the DSP chip. Electronic control unit (ECU): It is a centralized processor unit in a vehicle for controlling vehicle functions in real time. Electronics manufacturing suppliers (EMS). Electric vehicle (EV): An electric vehicle (EV), also referred to as an electric drive vehicle, uses one or more electric motors or traction motors for propulsion. High level operating system (HLOS): Operating system with the ability to run thirdparty applications. In-vehicle infotainment (IVI): Consists of information and entertainment: mobile phone and radio systems installed in vehicles, designed primarily for voice communications and products that provide in-vehicle entertainment to drivers and passengers, including audio and video playback, and Internet access. Includes both devices for in-vehicle playback of content (CD, DVD, MP3) as well as receiving and playing back content from external sources (radio, TV, satellite, Internet). Long-term evolution (LTE): A standard for wireless communication that supports high-speed data for mobile end user devices. It is based on the GSM/EDGE and UMTS/HSPA network technologies developed by the 3GPP (3rd Generation Partnership Project). Microcontroller (MCU): This is a semiconductor device that provides processing support for applications such as servo and motor control. Major difference between MPUs and this device is that it has some form of ROM, EPROM, or EEPROM, which stores 70

71 programmed customer supplied instructions. These instructions allow MCUs to carry-out control functions in various applications. Microprocessor (MPU): A microprocessor is the central logic-processing semiconductor that enables intelligence in a system. Non recurring engineering (NRE): The sunk cost of developing the product and not the production cost. Original design manufacturers (ODM). Original equipment manufacturer (OEM). Passenger and commercial telematics (PCTS): Passenger telematics; Products that provide navigation solutions, including GPS devices installed in the vehicle. Also includes automatic toll payment, vehicle financial contract compliance, vehicle health monitoring/communication solutions, and driver habit monitoring/reporting solutions. Emphasis is on automated information services for passenger vehicles. Also includes location-based services for consumers. Commercial telematics: products that provide navigation and fleet management solutions, including GPS devices installed in the vehicle. Also includes automatic toll payment, vehicle financial contract compliance, vehicle health monitoring/communication solutions, and driver habit monitoring/reporting solutions. Emphasis is on automated information services for commercial vehicles. Plug-in hybrid vehicle (PHV): A plug-in hybrid electric vehicle (PHEV), plug-in hybrid vehicle (PHV), or plug-in hybrid is a hybrid vehicle which utilizes rechargeable batteries, or another energy storage device, that can be restored to full charge by connecting a plug to an external electric power source (usually a normal electric wall socket). Real-time operating system (RTOS). 71

72 Table 11 IDC Taxonomy of Semiconductor Market Segments Definition of Semiconductor Market Segments Computing Wired Communication Wireless Communication Consumer Automotive Industrial, Energy, and Medical Defense and Aerospace This segment includes computer systems like desktop PCs, notebook PCs, PDAs, servers, workstations, mainframes etc., plus storage devices e.g. rotating computer magnetic storage systems, optical disks, floppy disk drivers, hard disk drivers, laser disk drivers, CD- ROM, CD±R/W, DVD-ROM, DVD± R/W, tape drivers, read/write amplifiers, read channel devices, mass storage pre-amps, etc. This category does not include dedicated music CD or DVD-related ICs, which are included under consumer applications. Devices that are usable under both, consumer and storage applications should be counted here. Lastly, this category also includes ICs designed for periphery applications like printers, scanners, monitors, keyboards, mice, etc. Includes enterprise and telecom service provider switching & routing, security appliances, 2G/3G/4G/LTE/WiMax access, backhaul, edge/aggregation and core infrastructure, broadband access headed and consumer premises equipment, TDM, VOIP/IMS, telecom optical network equipment. Consists of 1G, 2G, 2.5G, 3G, 3.5G, and 4G mobile phones, smart phones, mobile base stations for 1G to 3G, WLAN, Bluetooth, UWB, ZigBee, 2-way Radio, Cordless phones, etc. Refers to any device applications consumed solely for entertainment or media experiences including but not limited to game consoles, DVD players and recorders, HDTVs, portable media players, etc. Automotive applications including engine and power train controls, infotainment such as radio and video circuits, air bags, antilock brakes, active suspensions, navigation circuits, engine controls, display circuits, motor control, noise cancellation, etc. Industrial applications including factory automation; industrial electronic circuits; electronic circuits used in power/energy (such as solar, wind) for the generation, transmission and distribution of energy; medical applications including circuits used in portable and fixed medical devices. Defense applications and aeronautical applications including electronic circuits used in army, navy, and air force equipment, commercial aircrafts, and other associated applications. Source: IDC,

73 European Commission The Impact of the Introduction of Multicore Technologies on the Computing Market and Opportunities for Europe - Final Report Luxembourg, Publications Office of the European Union ISBN DOI: /

74 CATALOGKK EN-NUE NUMBER KK EN-N FR DOI: /10157 ISBN