It is both a pleasure and an honor for me to present the third annual Phil Kaufman Award to Professor Carver Mead of the California Institute of Technology. Before I speak about Carver and his many contributions to EDA and design, I would like to thank my fellow committee members, especially Wally Rhines of Mentor Graphics, and also Georgia Marszalek for her help coordinating a group of very busy people.
As I was preparing for tonight's award presentation and thinking about Carver's long and distinguished career in semiconductors and microelectronics, my thoughts drifted to the challenges we are all facing in the coming era of Deep Sub-Micron silicon technology-a topic that has been bothering me for quite a while now. The solution to the challenges of DSM must come from significant changes in design methodology-how we think about the design process-and certainly not simply from "better" sets of tools delivered around some extrapolation of the status quo. I was reminded, once again, of the fact that EDA is at least as much about design as it is about automation, as much about the creative process of organizing our thinking as about the tools and technologies we use to implement them. And so it is very appropriate, as we enter this new era, that the winner of the 1996 Phil Kaufman Award for "significant contribution to an advancement in the EDA industry" be someone whose contributions have emphasized design and design methods at least as much as the tools and technologies used to implement them-have emphasized redefining the problem rather than simply improving the solutions to the old one.
Carver began his career in semiconductors, then a branch of applied physics, back in the late 1950s with extensive work in the study of insulators and a phenomenon known as tunneling. By 1965, he had used his understanding and experience to build the first working GaAs MESFET.
Carver's career in what we now know as VLSI began in the late 1960s when he began to ponder the limits of silicon MOS technology. In 1969, he postulated that the important properties of silicon circuits actually got better as the circuits were made smaller. In fact, in this early work on what we now call scaling theory, he predicted our ability to build circuits at least as small as 0.25micron and perhaps 0.15micron and that "below 0.15micron its not so clear." Now this is important because if one scans the IEEE Transactions on Electron Devices, in 1979-a decade later--a fundamental limit of 0.5micron was predicted, in 1984 the fundamental limit was projected to be 0.2micron and by 1988, according to one group of authors, it had jumped back up to 0.4micron! Well, in spite of the theory, we know today we can build working devices with channel lengths of around 0.05micron and that the factor limiting size will be oxide reliability, the same factor predicted in Carver's work more than a quarter century ago! To me, this example illustrates a number of aspects of the man we are honoring tonight. First and foremost, his unfailing optimism! He was sure we could get there, and in an industry that has rewarded optimism time and time again, time has proven him correct.
In addition, he used his newfound insight to lead him to an entirely new level of research and development, a characteristic he has demonstrated throughout his career. By 1970, Carver was teaching his first courses on-and I quote-"VLSI Design" and was building his first multi-project chip. I quote from an interview published in Silicon Destiny: The Story of Application Specific Integrated Circuits and LSI Logic Corporation, by Rob Walker, where Carver said "By 1969 it was clear to me that some day we were going to be able to put millions of transistor on a chip." Carver, you were right! What scares me a bit, a quarter century later, is the possibility we may actually be able to put a billion transistors on a chip by the end of the next quarter century!
Carver went on to say, "At that time...I needed to figure out (a) What kind of thing do you do with million transistor ICs? And (b) How would you ever get the design right?" I think we've answered the first question, and we're still working on a satisfactory answer to the second one!
So how has Carver contributed to answering these two questions? I'll begin with the first one-What does one do with a million transistors? I'm sure everyone here, and a significant fraction of the rest of the world, has heard of the book Introduction to VLSI Design, by Carver Mead and Lynn Conway, first published in 1979. I certainly don't have enough time to talk about the many contributions made by Carver, his student, and his colleagues throughout the "Mead-Conway Revolution," as it came to be know, but one of the major factors was the effect his work had on the field of integrated system design-what we call System On a Chip today. By hiding the physics of silicon, by providing a set of abstractions and so an understanding accessible to those who had previously only worked with applications, software and systems architecture, Carver played a key role in bringing system design and silicon together. I well remember DARPA-sponsored meetings in the late 1970's and early 1980's where the young Dr. Jim Clark discussed the design of his first geometry engine, a direct implementation of his graphics research in silicon that would lead to the formation of Silicon Graphics, to mention just one example. I am also sure that the single-chip RISC concept would probably not have emerged as quickly or as dramatically as it did were it not for Carver's pioneering work, especially in the area of structured custom design. Making silicon accessible to system designers, leading to a set of abstractions and methodologies that were to influence a generation of EDA technologies, is certainly one of Carver's greatest legacies. These are the people-these system designers-who answered the first question for us, and we are counting on them again with our billion transistor problem as well.
Well, the second question--How would you ever get the design right [with a million transistors]?--led Carver to his work on design methodology and tools. To quote Carver, "...even if I had a program that would wire together a bunch of logic gates...just getting a logic diagram with a million working parts is a big deal...and [a conventional approach] would preclude me from taking advantage of the things that made MOS technology so attractive, such as precharge, pull downs, and everything one could do if they know the environment in which the circuit lives."
Carver's first IC design was a PLA-based finite-state machine, which is interesting in that he "invented" the circuit structure as an extension of his thinking about methodology (not knowing that TI and HP Fort Collins had invented the structure independently at about the same time), and because he generated it automatically from a program, using an "artwork language" Again, to quote Carver, "This was the first silicon compiler; it was great fun." It was great fun.
The silicon compiler concept, most well-known in it's Bristleblocks embodiment that was developed by Dr. Dave Johannsen and using Dr. Ron Ayers ICL (Integrated Circuit Language), was to infect a generation of researcher and form the basis for a number of start-ups, from VLSI Technology to Silicon Compilers Inc., Silicon Design Labs, and Seattle Silicon to name just a few examples. Many ASSP companies, like Cirrus Logic, were also strongly influenced by this work. And while the concept of the silicon compiler as an overall approach did not stick--or at least has not stuck yet!--it has had many influences on today's EDA industry. Module generators, or "compilers," for regular structures, like memory and datapath-for example, the recent products by Silerity--and extension languages-for example, Cadence's SKILL language was inspired by the work of Carver and his students-are a part of this legacy.
Of course, it was the silicon compiler that was to bring Carver and Phil Kaufman together. It was when Phil was President of Silicon Compilers Inc. and Carver was a member of the Board of Directors that they worked together promoting the Genesil technology. And while they certainly had their differences at times in how best to approach the task, I'm sure Phil would have been pleased to see the importance of the work being acknowledged here tonight.
When Amr Mohsen, one of Carver's Ph.D. students and the founder of both Actel and now Aptix, was formulating his approach to his new FPGA, he spent many hours discussing his plans with Carver and the first Actel arrays were "compiled" automatically.
It was Carver's "million transistor" thinking in the late 1960s that led him to the understanding that ultimately wiring, and not transistors, would be the main problem in microelectronics design, a fact that influenced not only his thinking about EDA and led to the silicon compiler as an efficient way of optimizing interconnect, but also his thinking about design and design methods. His work with T. M. Lin on wiring delay in RC networks, published during the early 1980s, has also been applied and referenced extensively in EDA research and development in recent years.
Well, here we stand on the brink of the next microelectronics revolution, with 100 million transistor chips by the end of the decade and working devices that will ultimately support more than a billion switches/chip. Systems on silicon, precharged dynamic logic, wiring management, pass-transistor logic-or Domino logic, if you rotate the transistors 90 degrees--self-timed building blocks and automatic power management, design portability and design re-use. Now where have I heard all these terms before? When Wilf Corrigan, the founder of LSI Logic and pioneer of the commercial ASIC industry speaks of "Object Level Design Methodology," where "designers would select a processor, RAM, ROM, and special logic with just a few commands" I am reminded of the silicon compiler. Personally, I would not be at all surprised to see procedural design-silicon compilation--play a central role in the next generation of whatever the ASIC industry evolves to become, along with many other ideas Carver conceived and evangelized throughout the 1970s and 1980s. And, Carver, I'm sure it will be "great fun" too!
As I think about the more than thirty years Carver has contributed to the electronics industry, a clear pattern emerges-from his early studies of the properties of insulators, to the invention of the GaAs MESFET. From his work on the limits of device scaling and the almost unlimited potential of silicon circuits, to the need for structured-custom design methodologies, to the silicon compiler. From VLSI, to the efficient implementation of CCDs, to neural networks and the artificial retina. One step leads to the next, each one a giant leap, one body of work providing him with the insight to envision and tackle the next frontier.
As an industry, we are very fortunate that Carver selected microelectronics as the basis for his career-his broad vision and creative insights, his unshakable optimism, his inspiration and his leadership, have had an incalculable impact on us all , as well as on our industry. I'm sure we will still be hearing the echoes of his insights for decades to come.
But perhaps it was actually the other way around, perhaps it was microelectronics that selected Carver, for I cannot think of another technical medium that would have been able to provide him with the challenges and opportunities a man like Carver needs to survive! They are still there, Carver, and we still need your help!
So, in summary, on behalf of Mrs. Kaufman, EDAC, and all present here tonight, I am both pleased and honored to be able to present the 1996 Phil Kaufman Award to Professor Carver Mead, a man whose leadership and inspiration about everything from devices to design has had a major effect on the EDA industry, through his teaching, his research, and his work directly with industry. He has certainly inspired many generations of professionals in our industry today.