presence here is indicative of the international importance of

Transcription

1 #4319Y Draft #5 - F SUPERCOMPUTER SEMINAR Robert M. Price October 19, 1983 I. INTRODUCTION Good morning. First of all thanks to each of you for being here. In view of your busy and demanding schedules, your presence here is indicative of the international importance of our topic. The application of supercomputers to problems of vital importance is growing at a more rapid rate than at any time in the history of our industry but along with that there has also been a growing concern as to the adequacy of our national effort with regard to their development. Control Data deeply shares that concern. The benefits of a forum such as this seminar, to examine the "state of the art" generally and at Control Data in particular seemed clear to us. And your presence a broad cross section of government, academia, and industry confirms it. This group today includes more than 40 organizations with almost an equal distribution among the three sectors I just mentioned.

2 R M Price CDC speeches Charles PAGE Babbage 2 Institute < For 26 years, despite the many fads in the computer industry, Control Data has steadfastly pursued a course of developing large computers for scientific computation. That was true when Wall Street abandoned scienfitic and engineering computation as a subject of economic importance, when the industry itself seemed to say that the answer was to be one line of (necessarily compromised) computers to meet both commercial and scientific needs, and even now as companies chase the "micro skirt." Well, you can imagine then how happy I am to report that Control Data's computer mainframe business has never been better. This is true whether one looks at backlogged units on order, revenue, or profitability. With all this in mind, it seemed appropriate to title my remarks this morning, "An Odyssey of the Elder Child." The large systems business is, of course, the original and oldest of Control Data's four businesses. And certainly the odyssey of the last 26 years has been a challenging and exciting journey. At times I know some of you have looked upon it not so much as an odyssey as abject wanderings, but the maze of strategic choices over the years has truly helped strengthen our commitment to be a leader in the supercomputer business. So, although the methodology we apply to stay on the fast track of advanced technology could vary with time, the objective will not.

3 PAGE 3 The basis for Control Data's large to super-class systems strategy is founded upon four straightforward principles: 1. "Focus on what we do best and do it better." 2. "Support what we sell and sell what we support." 3. "Leave no (technology) stone unturned." 4. "There is no substitute for preserverance." First, let's take "focus." As I said a moment ago, during these past 26 years we have continually emphasized the maximum "performance per dollar" for scientific and engineering processing. This focus still exists today. From time to time, we have tested new markets. Upon some of these occasions, we found success. But in many instances we discovered that trying to be an "all things to all people" company was driven more by youthful zeal than wisdom. This quest reached its peak during the late 1960's and early 1970's. Some years ago, we somehow developed enough wisdom to adopt computer product strategies which were based upon industries and related application areas in which the need for our particular products was greatest. Since that time, we have sharpened the focus on these "targeted industries" and support them on a priority basis in terms of marketing and support effort as well as R&D. (Flexibility,

4 PAGE 4 however, is a key Control Data characteristic and we certainly have been known to take an order or two outside these targeted industries.) Since several and others will be covering the subject of specific supercomputer applications in great detail during this seminar, there is no need for me to elaborate except to note that it appears all these targeted applications are represented by the attendees here today. For the future, we believe that through sponsorship of research in a variety of scientific, mathematical, and engineering disciplines, we will not only expand and refine the existing application base but will undoubtedly add some new ones as well. I'll come back to this point at the end of my remarks. Now to the second principle for success in the large and super-scale systems business "fully support what you sell sell only what you support." In the beginning Control Data's limited resources pretty much dictated a bare bones approach that is, we were a supplier of little more than high performance computer hardware. Over the years as we continued to reinvest in the business, we have grown to be a total system supplier with products ranging from

5 PAGE 5 high-performance local networking with super-class computational facilities to a full range of on-line hosts supporting graphics, file management, interactive time-sharing and learning centers. Products without this full complement of support services cannot truly meet your requirements. Upon occasion, Control Data's getting all of the pieces in place has quite possibly tested your patience. And truly we appreciate your support over many years. But today I can assure you that the support capabilities for Control Data's products are second to none. The third principle, "leaving no (technology) stone unturned" has certainly taken us down some blind alleys both in terms of basic technology and implementation. But as a result, we have sifted through a lot of options and consequently during the past five years we have established ourselves as the leader in the application of high-performance chip technology. Many of the so-called technology fads and reported breakthroughs in the past were suspect and dropped by Control Data's technologists long before their "public" demise. The strengths we have developed in this regard will assure further leadership in high performance technology. On the fourth principle of "perseverance" what more need be said. All we have to do is look at the record.

6 R M Price CDC speeches Charles PAGEBabbage 6 Institute < So instead of reviewing that, let me take a few minutes to talk about the future. As many of you know, Control Data has had a history of promoting two key concepts: first fostering technological cooperation and second helping to spawn new business ventures. As an industry matures, cooperation is the only answer to geometric increases in the cost of developing new technology, and many new ventures are the only answer to the problem of maintaining a high rate of innovation using that technology. During 1983, we have put in place two significant hi-tech companies. They were Microelectronics and Computer Technology Corporation (MCC) and ETA Systems. By now I assume most of you have collected sufficient information on the objectives of MCC. So I will limit my remarks on MCC to just a few words. MCC is a joint venture of 12 major U.S. companies in a ten-year program to do research in computer aided electronic design software (ECAD), advanced computer architecture, and decision theory for artificial intelligence. ETA Systems was founded just a few weeks ago on the first of September. Possibly it may be less understaood by some of you as to the reasons for its formation, its major objectives, and its relationship to Control Data.

7 PAGE 7 Later today, Larry Jodsaas will present more detail on the actual working relationship between Control Data and ETA Systems and in particular how all this relates to current CYBER 200 products and customers. Tomorrow Lloyd Thorndyke, with Neil Lincoln and Tony Vacca from his executive technical staff, will spend most of the morning walking you through the ETA organization, covering the primary objectives, technology strategies, and product objectives. But let me give you an overview on ETA. The business purpose of ETA systems is very simple. ETA is to design, manufacture, and market the world's fastest scientific computers. It's initial product mission will be to develop a ten gigaflop machine for delivery by the end of In order to do this, the architecture, VLSI technology, and software that Control Data has been developing under the CYBER 2XX project for successor machines to the CYBER 205 will be transferred to the new enterprise. Thus, initially the technology will flow from Control DAta to the new company. We will also provide at arms-length technical and management consulting services, design services, and possibly some administrative services. This will allow ETA to concentrate its financial resources and people on development of the new product. Control Data will

8 R M Price CDC speeches Charles PAGEBabbage 8 Institute < also make a commitment for a certain number of the initial machines. That is the simplest way we know to say that ETA has Control Data's full financial backing. It is nothing more or less than that. One of the major secondary benefits of supercomputer development at Control Data has been the flow of technology from that group to the other computer product lines. Through cross licensing arrangements with ETA, we will continue to enjoy such benefits. In fact, we will undoubtedly enjoy an increased and long-term technological benefit. And so the ultimate beneficiaries of this will be Control Data's customers, present and future, from supercomputers to the bottom of the line. I would like to close my remarks this morning with special emphasis on one particular point. Through its 26 years, Control Data has been a leader and an innovator in many regards. But perhaps the most important of these has been fostering technological cooperation. We have pursued that course long and vigorously from simple one-on-one cooperation in circuits and architecture to undertakings such as MPI which became a billion dollar enterprise to MCC which will become a major source of U.S.

9 PAGE 9 microelectronics and computer technology. Because of these undertakings, not only has Control Data survived and prospered, but literally thousands of small companies have access to more affordable technology and products. With ETA we have taken yet another step one which will not only help assure Control Data's future in supercomputers and help assure our nation of leadership in an arena of vital national importance, but also will set an example for large company/small company cooperation for all industries to follow. Further to this point, we feel very deeply the enormous need for cooperation between industry, government and universities in learning to better use the power of vector and parallel processor super computers. During the past five years Control Data has been sponsoring research within universities in the development of vector numerical algorithms. Tomorrow John Lacey will mention some of the results of this effort. The work of the National Science Foundation which will result in additional funding in 1984 clearly shows that something good is happening research programs which Control Data sponsored can now be continued and expanded by NSF so that we can pursue other avenues of assistance.

10 PAGE 10 With this in mind, we are developing a new Control Data sponsored "Program for Advanced Computing in Engineering and Research" PACER. PACER will result in some $8 million of additional research monies being directed specifically to applications of vector and parallel processing computers. More precisely, the program will include funding by Control Data of $500,000 for each of the next five years, to be matched by twice that amount from other sources. One feature of the program will be to cooperatively support fellowshios for graduate students in the development of super computer applications. We are open to receive your thoughts at this meeting as to the needs and the role such a program can play. Details will be released before year end. So, in a very real sense, we have put our money where our mouth is. But more than that, we share with you the challenge and the excitement of what the world's most powerful computers can accomplish. Thanks for being here and I hope you have a rewarding seminar. Any questions?