Oral-History:Federico Faggin
About Federico Faggin
Federico Faggin was born 1 December 1941 in Benito Mussolini’s Italy. Intellectualism ran in Faggin’s blood. His father was a teacher in the history of philosophy and general history. To his father’s dismay he was interested in electronics, not the humanities. At a young age, Faggin realized that his interests were the opposite of his beloved father. He remembers being interested in machines and anything mechanical at a young age. Even as a young boy, Faggin recalls the irony of the lure of technology: he felt he could understand why machines worked, but not humans.
At about the age of eleven, this young man saw a gentleman with a model airplane. He was captivated. “Wow,” he said with only a child’s awe and admiration, “it’s flying.” Inspired, he rushed home and built his own model plane. When it was completed, he took it back to the park with visions of seeing his masterpiece take to the sky like the birds that had captured his attention. It crashed. At that moment, the young man who had stimulated this whole obsession happened to be riding past on his bike and took pity on this little boy he perhaps had never noticed beforehand. He showed him how to build the plane, explained what would make it fly, and why his had not worked. Faggin associated flying with birds, and thought it was amazing that humans could create something that could soar like them. He studied electronics at his technical high school because he wanted to design and build airplanes, but the school had gotten rid of aeronautics program.
Faggin was hired as an assistant engineer in Olivetti, Italy in the mid 1950s. He recalls always being fascinated by physics. In his opinion, it really explained how and why things worked, which, from childhood, was what most interested him anyway. The company he worked for in Olivetti put Faggin in charge of building a computer. It ended up being seven feet high and as wide as a door frame. Even though he was now building machines, he felt something was lacking; he needed a formal education. Faggin enrolled at the University of Padua, in Italy, completing a program that usually took five-to-seven years in four, while working and giving private lessons to boot.
Even though he mastered physics, Faggin remained interested in application and the building of machines. He graduated in December 1966, receiving his degree on his 24th birthday. Aftewards, he would teach in Italy for an entire year. When interviewed later, Faggin believed that learning the English language and studying in both Italian and English gave him an edge over many of his European colleagues. He eventually came to America and took classes at Stanford in the late 1960s. He also got a job in a small start up company in Silicon Valley by the name of General Micro Electronics in the summer of 1966. Faggin was making his first computer in the early 1960s but did not realize it then. In February 1967 he joined SGS Fairchild. It was easier for him to work in the US than in Italy, because of the emphasis on production and the politics of the industry. In the U.S. if you produced you were in. In Italy, one could sustain a position if they played their politics right.
After a short while at Fairchild, Faggin developed the silicon gate technology, which was the first process technology that allowed practical manufacture of the self-aligned gates. The people who formed Intel took his technology and formed their mega-company, to which he was given zero-credit. They basically stole the technology from him and Fairchild. Intel undermined Fairchild, not only stealing technology, but key employees as well. In light of these developments, Faggin decided it was time to leave Fairchild. He called his former boss who was now working at Intel and asked if he had work for him. A few months later he received an affirmative reply and was soon working on a custom project that required a lot of logic. They had been sitting on the project for almost six months, and the customer was coming to check on the project’s progress in a few days. After working for a series of high-powered companies Faggin decided to start own company, Zilog, and took Intel head on, which was backed by IBM. Faggin agrees with Kleinrock about technology and that we are in the beginning of something that is going to get even bigger.
About the Interview
Federico Faggin: An Interview Conducted by John Vardalas, IEEE History Center, 27 May 2004
Interview #442 for the IEEE History Center, The Institute of Electrical and Electronics Engineering, Inc.
Copyright Statement
This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Director of IEEE History Center.
Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, IEEE History Center, 445 Hoes Lane, Piscataway, NJ 08854 USA or ieee-history@ieee.org. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user.
It is recommended that this oral history be cited as follows:
Federico Faggin, an oral history conducted in 2004 by John Vardalas, IEEE History Center, Piscataway, NJ, USA.
Interview
Interview: Federico Faggin
Interviewer: John Vardalas
Date: 27 May 2004
Place: Faggin’s Faubion Office
Childhood interest in science
Vardalas:
When and where were you born?
Faggin:
I was born in Italy, in Vincenza, December 1, 1941.
Vardalas:
What are your earliest recollections, as a child, of your fascination with science? Can you recall when you first got interested in anything scientific?
Faggin:
Yes, I was very much interested in machines as early as I can remember: airplanes, cars, bicycles—anything mechanical was really appealing to me. In fact, the fascination with machines was because I could understand how a simple machine worked, where people, I couldn’t!
I remember that one of the best gifts that I was ever given was a Meccano when I was nine years old. I really enjoyed the Meccano so much because I could build things. I wasn’t following any instructions; just my imagination. It was very empowering. The pieces of the Meccano are about one centimeter wide and they’re made of steel, but the Meccano that my father bought for me was actually made of aluminum and it was about half the normal size of a Meccano. You could build machines as big as you wanted, but they were much nicer compared to the original because the Meccano, which is a bit unrefined. And it had very precise gears and all kinds of wheels of different sizes.
Vardalas:
Did you find yourself doing Rube Goldberg-type devices? Or were you trying to make devices that were practical examples of things out there?
Faggin:
The first thing that I made was a wheel with rainbow colored sections that became white when you spin it. I had seen an illustration of it in my schoolbook, so I actually made it. I could see it turning white! But what I liked to construct most was little cars and trucks. It was a pretty rich set with enough pieces to allow me to make many different items together, without having to disassemble some of them to recover the pieces.
Vardalas:
Did school reinforce this interest at all? Or were they separate interests?
Faggin:
No, certainly not the elementary and middle school, but I attended a technical high school, which definitely reinforced my interests.
- Audio File
- MP3 Audio
(442_-_faggin_-_clip_1.mp3)
The other experience that was extremely important for me – it happened when I was 11 – was seeing a model plane. There was this young man in the middle of a field where I was playing with my friends. He was holding a neat model plane; he wound up the propeller, let it go, and it took off! It was just unbelievable to me; I thought that only real airplanes could fly and here this model Piper Club airplane was soaring! I was running after it, captivated. “Wow! It’s flying!”
And then I decided I was going to make myself a plane like that. At home, I collected sticks and other useful materials, and got the modeling paper that my mom used to make dresses with, and I made one. Of course, it violated every single law of aerodynamics, and it was poorly constructed, as you can imagine. I went into the same field where I first saw the guy flew his plane, which was just walking distance from my home, and of course, and I launched my model and it just crashed. At this point – a lucky coincidence – the same fellow was coming by on his bike. He stopped and asked me to come over and let him see what I had. I showed him my contraption, and he started laughing. But he kind of sympathized with me and started explaining how to build it right, where to get the construction material and so on, and where to buy a book about model planes. The first book that I bought myself with my own money was a book on how to build model planes.
In fact, that was absolutely the most formative experience of my youth. I soon built a second model. That one didn’t fly either, but the third one flew. I was designing and building the whole model from scratch because I didn’t have money to buy kits. So I was purchasing just the bare-bones supplies, sheets of balsa and covering material. I even made my own glue by dissolving celluloid in acetone. That’s a passion that I had for all my life. I still go fly to this day. You see that I have a good color in my face? That’s because on weekends I go flying my own radio controlled planes!
Vardalas:
As a child, then, when you got into this, were you captivated at all by what made these things stay up? Or were you more interested in how to make it work and how it looked?
Faggin:
The whole thing. I think it was more of a gestalt for me because it was the idea that something that was made by a human being could fly like a bird – and the grace of it. There was also an aspect that was almost a ritual about the whole creation process. By making model planes, I basically learned – without knowing that I was learning – the complete product cycle. I imagined something planes the complete cycle of imagining something, making a plan, getting the material, building it, testing it, and enjoying it, the full cycle from conception to making it work. By the time I was 12, this entire creative cycle was routine for me. That’s why all my life I have been quite good at realizing and building things.
Vardalas:
You must have come into the conflict of form versus function.
Faggin:
Oh yes, of course, though I didn’t perceive it as a conflict. I wanted by model to look good, but the first priority was that it would fly well. And then because of my passion, I chose to go to a technical high school that was teaching aeronautics. Unfortunately, they had stopped that specialization, without my knowledge, so I had to choose a different one and I chose electronics.
Vardalas:
So that’s the transition that, then. You originally wanted to do aeronautics?
Faggin:
Yes, that’s what I wanted to do. I wanted to design and build real airplanes.
Vardalas:
Your life would have been very different if they would have had that option.
Faggin:
That’s true!
Vardalas:
Some children or some young adults would have -- depending on their curiosity -- would have said, “I’m more interested in learning about the science of the planes, why they stay up.” And that is, of course, science-based engineering. Were you interested in that?
Faggin:
Not really. Just designing, building and making them fly. The pleasure of seeing them fly for the first time was the peak experience. At that time, the underlying science was important only if it would allow me to design better planes.
Vardalas:
The thrill of having it work.
Faggin:
And having it work, yes. Absolutely.
Family background
Vardalas:
Let me ask you about your family. How supportive was your family? You mentioned your father bought this Meccano kit for you. What did he do? Did he do anything that would have incited you to have an interest in this area?
Faggin:
No, he didn’t, he bought me the Meccano because I was so fascinated with machines. He understood that that would have been a great instructional toy for me. It was more than a toy, as a matter of fact; it was something that would allow me to express my creativity instead of playing with it for ten minutes and breaking it.
But my father was a teacher of the history of philosophy and history, in the Classical Lycee. He wrote many books and papers and he later also taught courses at the University of Padua. Most of his life he taught at the “Liceo Classico,” a very formal and, at that time, very elitist school where you study Greek and Latin more than you do the sciences.
Vardalas:
Okay. I get it. Like the public schools in England.
Faggin:
Yes, like certain public schools. Now, I had no desire to go to this school, to the chagrin of my father, because I wanted to go to this other technical high school, which has a very different, more practical curriculum. It had a five-year program with 44 hours a week of laboratories and lectures, plus much homework on top of it. It was a very demanding school.
Vardalas:
It was okay for your father, being a man raised in the humanities? There was no snobbishness about you going into mechanical things?
Faggin:
There was a little bit. There was a little bit.
Vardalas:
Did you have any siblings? Any brothers and sisters?
Faggin:
Yes.
Vardalas:
Did they follow a similar path?
Faggin:
No, I’m the only “mechanic” in the family. I have a brother, two and a half years older than me, who followed in the footsteps of my father. Not so much in philosophy, but humanities and art. He’s actually a specialist of Flemish art. But he also studied philology and languages; he wrote a dictionary of Italian-Friulano, which is a language in the northeast of Italy. Of course, Friulano has the status of a dialect right now, but it has its own literature. He also translates from Dutch into Italian. Because of his Flemish art interest he spent a fair amount of time in Holland. He is now a professor of Dutch art and literature at the University of Padua. My youngest brother, the youngest of the siblings, is a practicing lawyer. My sister, who is right after me, was a fashion designer; she taught fashion design and drawing and now she teaches history of art in high school.
Vardalas:
So in a way, she’s like you a little bit. She went into the product creation.
Faggin:
A bit different, more like my father, who also had an artistic vein. He used to paint, particularly surrealistic, intellectual watercolors in the vein of Magritte. He was quite good, actually. So, in the end, each of us took some aspect of his interest.
Vardalas:
Sounds as if that influenced you, the creative part of your father, the painting?
Faggin:
Yes, perhaps the painting. But I also benefited from the cultural environment in our home and absorbed his love for literature and music, and especially his intellectual integrity and inquisitiveness, qualities that I developed more and more as I grew older.
High school; employment at Olivetti lab
Vardalas:
Tell me about this high school. You went there for aeronautics.
Faggin:
I went for aeronautics. But because it was not being taught anymore, I chose the second best direction, which was radio technology and electronics. I reasoned that this way I could design and build my own radio-controlled plane [laughs]. As you can see, I had a one-track mind! That was part of my initial motivation but, of course, I also liked very much learning new things. My interests were not so limited, for example during those years I developed a love for Russian literature and I read all the major works by the classic authors.
That school was tough, but I did very well. I graduated by far the best student of the entire school with a grade-point average of A, when the next best student had an average of B. I am referring to the final exam, which is a state exam given by a board of examiners from other Italian schools, at the end of the required five years (the high school in Italy lasts five years, not four). Then, after graduation, I went to work for Olivetti.
Vardalas:
Directly from high school to Olivetti?
Faggin:
Yes. Most of the designers of the Olivetti business machines came from similar technical high schools. They were not graduate engineers as it would be commonplace in the States.
Vardalas:
Oh, I see. Before we can pursue that, did anyone in high school, any professors or such, play a prominent role in your life and shape it?
Faggin:
There was one professor who played an important role – more than I realized at that time – my teacher of physics. We started studying physics in the second year of high school. It was fascinating, because I could begin to actually understand the deeper reasons why things work the way they do. He made me appreciate the subject, the fact that there was a body of systematic knowledge teased out of nature over the centuries by curious and bright people. In fact, I eventually took physics at the University because of my desire to go deeper yet than I had done before. Previously I had learned simplified theories of operation and how to make and use things, now I could learn how they functioned based on first principles. How did a transistor really work, for example, beyond the pragmatic understanding of its operation, sufficient to put it to practical use? Of course, I later found out that if you keep on asking probing questions, you still find out that nobody knows!
Vardalas:
So you finally found out why your planes worked?
Faggin:
No. That much I knew from the book I bought when I was 11. That book was well-made, and it had a fair amount of aerodynamics and the theory of all the forces operating in a plane. I didn’t understand everything because it dwelled quite extensively on theory, but I certainly understood the basic theory of flight.
Anyway, I went to Olivetti to work in their R&D laboratory, which had been set up only a couple of years before. It was 1960.
Vardalas:
What city was this in?
Faggin:
It was in Borgolombardo, near Milan. The original laboratory was located in Pisa, as an adjunct to the University of Pisa, and then moved to Borgolombardo. Curiously, it all started by instigation of Enrico Fermi. In the mid 50’s Fermi visited the University of Pisa and recommended that they get involved with electronic computers -- commercial computers had just begun to be available. Out of that suggestion, and of course other considerations, came this experimental laboratory, financed by Olivetti. After putting the team together and doing some initial work, Olivetti decided to create a permanent R&D laboratory. That was the one in Borgolombardo.
I was hired as an assistant engineer, which was essentially the position you got when you came out of my type of school. But I was judged a bright kid so within a few months I was given the assignment of actually designing and building a computer. It was a computer about as wide as a door frame and about seven feet high. Racks, card cages and circuit boards. It was basically a decimal computer where the accumulator was made out of ring counters. It was an adaptation of an American concept published in a technical magazine. The idea was to scale it up and build an experimental computer and see what you could do with something like that. This computer had a 4K-word magnetic core memory. A word was 12 bits. I designed all the central control and built the machine
Vardalas:
Where did you get your understanding of digital concepts?
Faggin:
I took a course at Olivetti at the beginning of my employment. It was about one month long. Then I bought some books and learned from them
Vardalas:
Right on the job?
Faggin:
Right on the job. Then, the engineer I was reporting to and directing me, got into a car accident, so he was off work for three months. Out of necessity, I took over. I started doing things on my own and by the time he came back he left me alone because I was doing so well
Vardalas:
I’m interested in -- within organizations -- when they’re on the beginning of a learning path, how they encourage experimentation and learning. In some companies, people are thrown in and they’re not really encouraged. Were you left alone? Was there encouragement? How did they foster this idea of try, fail, but we’ll help?
Faggin:
I don’t think there was an explicit or deliberate system at Olivetti. It basically just evolved that way. Mine was a project that didn’t get a lot of attention; it was not in the critical path of the lab’s work. It was not intended to be a product; it was basically an experiment. They wanted to see how it would work and whether they could build a small computer. The core activity of the laboratory was to develop a big mainframe. My activity was part of what was called the Circuits Lab. My project was a little bit of a bootleg project of the Circuits Lab, to learn more about circuits. So I designed the transistorized circuits, I designed the logic, I built it, and I got it to work in about a year.
When I got there, the project had just started. Some work had already been done, but very little. I basically completed the project.
Vardalas:
What was the fate of this project?
Faggin:
With my leaving, it was basically left alone. I was told that it actually inspired the Programma 101, a small computer that was very successful. The idea was to miniaturize a computer of the same general capabilities of the one I made, although the architecture was also changed. The idea that you could have a computer with memory and everything else, all fit in one rack, was not bad for that time, of course.
At that time I used germanium transistors. They had a storage time of one and a half microseconds! Now you can run an entire software routine in the time it took to switch one transistor! In 1960, the much faster silicon transistors were just becoming available at Fairchild. Fairchild Semiconductors was founded in 1957 or 1958, and their products were extremely expensive early on. 1960 was right at the forefront of the wave of change that quickly made obsolete transistors for computers and replaced them with bipolar integrated circuits.
Vardalas:
Was this a difficult experience for you? Was it exhilarating?
Faggin:
It was absolutely fun. I had five technicians working for me, so I was actually managing the entire project. And again, I could not have done it that soon in my career, had I not internalized how to run a project from my experience in making model planes. Instead of designing and building a model plane, I was designing and building a computer, but it’s the same process.
Vardalas:
Did anybody come knocking on your door every month saying, “Well, Federico, what’s happening now? Where are you at?”
Faggin:
[Laughing] Obviously, I would routinely report on my progress. I wasn’t just left alone in a corner. But my boss would occasionally be pleasantly surprised by the progress that was going on. I worked intensely.
Studies in physics, University of Padua
Faggin:
Anyway, as I was completing this project, I decided that I wanted to go back to University and study physics.
Vardalas:
Why not electrical engineering?
Faggin:
Because I didn’t need more electrical engineering in my background. I already was practicing that stuff. I could obviously learn more, of course, but I wanted to get a deeper understanding of the foundations of science. And I wanted to get into solid state physics.
Vardalas:
Okay, I see. So you were already focused on--
Faggin:
No. I really had no interest in doing particle physics or any of that. But I wanted to learn how things work and how to invent new things, starting from the basic laws of nature.
Vardalas:
So would it be safe to say, then, that your interest in physics was rooted in this concrete technology?
Faggin:
Yes. At that point, I didn’t see myself as a scientist, as somebody who would spend his life studying laws of physics. I mostly wanted to learn, to gain a better understanding of the foundations of science and math so that I could be much more proficient at doing something similar to what I did at Olivetti. Contrary to many kids that go straight from high school to University, not really knowing what they want, I knew exactly what I wanted. But I was also genuinely interested to find out more about physics for its own sake. Just for the pleasure and the curiosity to know about the nature of reality.
Vardalas:
Which university did you go to finally?
Faggin:
University of Padua.
Vardalas:
What made you decide on Padua? Padua has a very long tradition.
Faggin:
Simply practical reasons. First of all, it’s one of the top universities in Italy. But also it’s close to home. Vicenza is only 35 kilometers away, about 20 miles. It’s a 20-minute train ride to the University.
Vardalas:
I see. Was your father happier now that you were going into academics?
Faggin:
Actually, it was very interesting, because when I told my father that I wanted to go back to school to study physics, his first reaction was, “Oh… You’re already doing so well. Why don’t you stay at Olivetti?” I didn’t expect that reaction and I didn’t understand what was behind his response. Did he think that I couldn’t make it? He had this idea that if you went to a technical high school, you may know practical things, but you may not be intellectually prepared for university. I thought, “Does he think that I cannot make it? Or is it because he doesn’t want to support me?
He was basically discouraging me. He didn’t say no, but he was a bit negative. I said, “Dad, look, all you have to do is give me a place to sleep, and when I’m home, give me food. I will take care of everything else myself: tuitions, books, travel, everything.” And I didn’t ask one penny of my family all my university years.
Vardalas:
And you worked right through university?
Faggin:
Well, I saved enough when I worked at Olivetti. Plus I gave private tutoring. So I never asked him for any money. I was able to take care of myself all throughout my school years. And I was the only student of my class that was able to graduate in the minimum allowed time, which was four years.
Vardalas:
Is that right? Usually it’s five.
Faggin:
Right, it’s five. Most people actually take six or seven. I did it in four years and got summa cum laude. Part of my motivation was also to show my father that he was wrong thinking that I was intellectually not up to the task – “I’ll show you!” [Laughs]
Vardalas:
The power of the father figure! What are your most vivid memories of your undergraduate days in physics? Does anything stick with you anymore about those years?
Faggin:
Well, my first year was very tough because I lost the first couple of months. I worked at Olivetti until the end of December, then I took the Christmas holidays off and I started going to school in January, when classes had started in October. My first lectures were a disaster, I didn’t understand anything the teachers were talking about, particularly in math, because they were already well into the program. I was forced to be in a catch-up mode for most of a year and a half, studying mostly by myself.
In Italy the system is different than here: you don’t have to attend the classes and you are not given homework that is checked regularly. At the end of the school year, you just take a written exam and if you pass it, you are admitted to the oral exam. If you pass that, you pass the course. I had to attend the laboratories, however, so I went to all the labs. But I missed the theory classes: physics, chemistry and math, so that I could dedicate myself completely on catching up. Furthermore, in Italy there’s no general education course work, you go straight to the subjects relevant only to the direction you’ve chosen.
So I had to do all that by myself. Math was by far the hardest.
Vardalas:
Were you having any doubts at this point?
Faggin:
No, actually no doubts. I was absolutely determined to succeed, but the first year was hard. Eventually, not only did I catch up, but then I started accelerating, pulling ahead of the class both in grades and coursework completion. My first few exams I had B-, B. Then I started having B+, and then after a little while, it was all A and A+, until the very end.
Vardalas:
Did you find yourself having a common outlook on things with your fellow physics students? Coming from a practical world—usually physics majors are not necessarily that oriented.
Faggin:
Early on I hooked up with a guy that, like me, had a diploma of “perito,” which means “expert,” having attended a technical high school like mine. He had also chosen physics for my same reasons and had worked as an assistant engineer at the European Nuclear Research Center. We basically helped each other during the first few years, but I interacted with a lot of other students as well and I never felt any different from them.
Vardalas:
I can assume, then, you were more interested in experimental physics than you were theoretical physics?
Faggin:
Well, I was quite good and interested in both. For example, I was the only student in my class to get “trenta e lode” which is basically the absolute maximum grade, it would be like A+ cum laude, in theoretical physics. So I actually was quite capable all around. But of course, while I enjoyed theory very much, my real desire was to apply it to designing and building things.
Graduation; assessment of educational experiences
Vardalas:
What did you do when you graduated?
Faggin:
I stayed at the University for six months. I taught electronics laboratory to third-year students – the advanced labs start with the third year. I also continued to work on my thesis subject – flying spot scanners. For my thesis, I designed and built a flying spot scanner for automatic reading of scintillation camera photos. I started by designing and constructing all the equipment for the experiment, from the vibration-free experiment table, the CRT focusing magnet, operational amplifiers to the digital control. I did not build the entire system but just the core portion of a flying spot scanner using a novel opto-electronic feedback technique for getting precision measurements.
Vardalas:
What year was this?
Faggin:
It was 1966 because I graduated in December, 1965.
Vardalas:
So you taught for a semester?
Faggin:
I taught the full academic year, starting in October-November 1965, right after I had finished all my exams and submitted my thesis, but before my thesis discussion, which is the final rite of passage. I graduated on December 1st, on my 24th birthday, and I continued teaching until June 1966, the end of the academic year. In Italy, the academic year goes from October until June. Italy does not have a semester system.
Vardalas:
At the time, were you satisfied with the course of study? Did you feel this is the way things should be? What were your attitudes towards the kind of education you were getting? In retrospect, what do you think of it now?
Faggin:
In retrospect, I think that the basic theoretical coursework was actually reasonably well done. But the material was not always all that updated. It was five or more years behind what was actually going on. In the US you learn about topics that are only a few years away from the state of the art. In some cases, we had very poor textbooks. One of the reasons I was quite proficient was that I did integrate my textbooks with English-language books used by American or English universities. Of course, you wouldn’t do this here, because it’s a different system, but even in Italy most students didn’t do it because it required extra effort and many of them didn’t know English. That’s when I started developing a real appreciation for the quality of the texts used in the Anglo-Saxon world. In this way I was able to overcome some of the limitations of the Italian system.
When I eventually came to this country, I had absolutely no problem fitting in. In fact, I felt that I had quite a good theoretical preparation. For example, in 1969 I took some graduate courses that Stanford University was giving at the R&D lab of Fairchild. At midcourse, I became an assistant to the teacher. Once you have a good grounding – and I believe that I did have a good grounding in math and physics – then the rest you can learn on your own
Vardalas:
Do you know Harold Rosen? He did the first geosynchronous communication satellite. He went to Caltech, and he said the best education he got was a sound grounding in math and physics.
Faggin:
That’s right. After I received my degree in physics, I could tackle any reasonable scientific problem – even ones I had not studied before -- because I could quickly find my way around the literature and figure out how to approach it.
Employment at CERES and SGS-Fairchild
Vardalas:
In 1965 you graduated, and then you worked?
Faggin:
And then I worked at the university until June, 1966. In July, 1966 I went to work for a small company near Milan that was making thin film circuits.
Vardalas:
Is it SGS?
Faggin:
No, CERES, just a small start-up, one of the rare high-tech start-ups in Italy. The engineer who began this company was my old boss at Olivetti. In fact, he was the guy who broke his leg in the car accident that led to my having to design the computer by myself! That little company -- a bit more than a half a dozen people -- was also the sales representative of GME, General Micro Electronics, the first pure-play company in MOS integrated circuits. GME was a start-up company in Silicon Valley, with the founding team coming from Fairchild Semiconductor, trying to make a business out of MOS integrated circuits.
I was hired to be the technical expert for the GME product line and for other projects as well. My first job, right after being hired was to go to California -- it was the summer of 1966 -- to take a one-week course at GME to learn about MOS technology and the GME product line, so that I could do my job.
My visit to California was another of those very powerful, life-defining moments. California was a different world. I was really struck by this place, where if you are an engineer, you’re not an alien, not to mention the incredible weather I found. Everything was so fantastic! I think at that point a bug started working in my mind, planting the idea that coming to America may be a good thing…
Soon after I returned to Milan, the University of Rome ordered a few MOS shift registers -- GME had introduced a 100-bit dynamic shift register, which was state of the art in 1966. But we never got delivery. They just couldn’t produce them. GME was one of those companies that are a lot of talk but little substance. In fact, to survive, they ended up making bipolar transistors, unable to produce reliable MOS integrated circuits. In those days, it was very difficult to make stable devices because of impurities in the gate oxide. Incidentally, it was the development of the silicon gate technology that finally allowed the production of reliable integrated circuits. That happened later, of course, and I had no idea I would be the one to do it in only a few years.
As a matter of fact, it’s all quite a coincidence because, in 1960-1961, I designed and built my very first computer, and in 1970-1971, only ten years later, in California, using the silicon gate technology, I designed the 4004 and the other companion chips that were much more powerful than the computer I built at Olivetti. And the entire computer fit in a single circuit board instead of a few hundred. Amazing! That decade was amazing! From transistors to a single-chip CPU in ten years!
Anyway, back to our story. GME was sold to Philco-Ford a few months after I returned from California and CERES stopped being GME reps because Philco-Ford already had their own Italian rep. So I ended up doing a number of different projects. But the company wasn’t really going anywhere. That’s when I joined SGS-Fairchild, the core of what is now STM, one of the top five semiconductor companies in the world.
Vardalas:
And they had an Italian subsidiary? Is that how you joined? Or you joined right in America?
Faggin:
No. SGS-Fairchild was an Italian company. It was 70% owned by Italians and only 30% by Fairchild. Olivetti, had 30% ownership in it and Telettra another 30%.
Development of process technology for integrated MOS circuits at SGS
Vardalas:
You mentioned in your interview with Astrid here that SGS -- and I assume you’re referring to the Italian firm SGS -- developed the first process technology for MOS?
Faggin:
Yes.
Vardalas:
What does that mean, developed the first process? When you say process technology, what do you mean?
Faggin:
What it means is that, when I joined the research laboratory of SGS-Fairchild, because of my knowledge of MOS technology acquired in that one week course I took in California, they asked me to develop the process technology for making MOS integrated circuits. So I developed the first MOS process technology for SGS – not the first MOS process in the world. Basically, SGS had never built an MOS transistor or an MOS integrated circuit before, so I figured out how to make them. Process technology means a method to fabricate integrated circuits.
Vardalas:
What did that entail?
Faggin:
Well, studying technical articles, reading papers and books, and doing experiments. Within six months, I actually had the process technology working, and I had designed two integrated circuits that became commercial circuits.
Vardalas:
Did you need funds for this?
Faggin:
Oh yes. I was working for the research and development laboratory, and it was fully funded. They also decided that they wanted to get into the MOS business as another branch of their business. About five months into it, I had two or three people working for me. So I had a small group, and that way we got the resources to get the technology developed.
That takes us to the end of 1967. Then in February 1968, I came here to California to work at the Fairchild R&D lab for a period of six months. By the time I left SGS, I already had those two MOS integrated circuits working.
Corporate culture of SDS Fairchild; comparison of labs in Italy and the U.S.
Vardalas:
What were your impressions when you wound up in SGS Fairchild here in California?
Faggin:
Oh yes, of course.
Vardalas:
What were your impressions when you first came here?
Faggin:
First of all, the R&D lab of SGS was nothing compared to the R&D lab of Fairchild, which was a big enterprise. I think they had over 400 people. It was a big operation with all kinds of departments, from device physics all the way to packaging technology. At SGS we had about 50 people – still a sizable group – and we had one pilot fab (wafer fabrication facility). The Fairchild lab had three or four different fabs – almost every major department had its own little fab where they would make special things with their own furnaces, their own reactors, mask aligners, etc. The amount of effort and also the quality of knowledge that was here was much higher than in Italy. In those days, Fairchild R&D Lab and Bell Labs were the two world leading laboratories in integrated electronics. SGS had hired a number of key employees with US working experience so the overall quality was not bad, but we were not a match for Fairchild. For example, SGS had Heikki Ihanthola, who was a Ph.D from Stanford University that developed one of the early device models for MOS transistors, working with Professor Moll.
Vardalas:
Was the corporate culture at Fairchild what you expected to find? Do you remember what you were stuck by when you came? Did you find what you expected to find when you showed up there? How did you find the corporate culture when you got there?
Faggin:
I think one of the first things that struck me was a bit like, “Here’s your project. Go do it.” Much more matter of fact here.
Vardalas:
Like, “No big deal. Go do it”?
Faggin:
Yes, go do it. They had a regular meeting every week, and there I would report what was happening.
Vardalas:
But they expected positive results. “Here it is. Now give us the answer.”
Faggin:
Yes, as a matter of fact. That for me was unusual, because in Italy they tended to be more roundabout and – even if was mostly appearance – more collaborative. Particularly my boss, Les Vadasz, epitomized this style. But my boss was a bit unusual too, because he was a native Hungarian who came in the States in 1956. So he did not necessarily represent the American way. He was a pushy kind of guy. In retrospect, my initial reaction was mostly a reaction to Vadasz, more than a reaction to the American system. But how could I distinguish anyway? My English was okay, but it certainly was not fluent, and Vadasz didn’t wear a sign saying he wasn’t American.
However, I fit right in without much effort, from the start. It was easy. Because again, I loved to do this kind of work. Work was so much part of what I enjoyed, what I wanted to do, and it was much easier for me to work in the States than in Italy. You know, in Italy, there are people who watch by the sidelines and don’t do much work. Out of ten people, three do a lot of work, two don’t do anything, and the rest is in-between. Somehow you cannot get rid of the two that don’t do anything because the system guarantees employment: unless they kill somebody on the job, you cannot fire them [Laughs]. In the US, you’ve got to produce; and if you don’t, you are asked to leave. Being one the three people that work hard, I think this is a much fairer system.
Development of silicon gate technology in microprocessing
Vardalas:
Speaking of producing, in that same interview with Bill Aspray, you said, “I developed the first workable processes that put memory on microprocessors.” Am I quoting that correctly?
Faggin:
No.
Vardalas:
How would you phrase it?
Faggin:
Within a short period of time at Fairchild, I developed the Silicon Gate Technology, which was the first process technology that allowed the manufacturing of self-aligned gate MOS integrated circuits. This was the technology that made possible the fabrication of semiconductor memories and the microprocessor. Self-aligned gate was a bit of the Holy Grail of MOS technology. It had been recognized from some time that to remove the limitations of conventional MOS devices, mainly due to the parasitic capacitances between the gate and the source and drain regions, one had to make the gate self-aligned, but nobody had done more than proof of principle before. Parasitic capacitances were primarily due to the overlap between the aluminum gate and the source and drain electrodes. This overlap was in turn required to compensate for alignment tolerance in the manufacturing process. The overlap capacitances, especially the gate-to-drain capacitance (which is increased by the Miller effect) degraded the speed dramatically. Before silicon gate, we had slow devices using a lot of silicon area, because they were metal-limited. Furthermore, they were unreliable because we could not clean up the system of impurities after the metal gate had been deposited. In summary, metal gate MOS integrated circuits were slow, unreliable and somewhat wasteful of silicon area.
Vardalas:
What was the solution you came up with?
Faggin:
The solution was to develop a process where the parasitic capacitances are minimized by eliminating the overlap between the gate and the source and drain regions. By making the gate out of polycrystalline silicon instead of aluminum, and letting its actual position define the source and drain regions, it was possible to solve this problem. In other words, the polysilicon gate acts as a mask against both the etching of the gate oxide and the doping of the gate regions during the doping of the source and drain junctions. The gate is said to be self-aligned because the definition of the source and the drain is controlled by the actual position of the silicon gate. Alignment tolerances will simply slightly change the source and drain dimensions leaving the overlap capacitances unchanged.
Vardalas:
All right. How did you come on this? How long did it take you to figure it out? Was this straightforward?
Faggin:
A week [laughs].
Vardalas:
[Laughs] Was it obvious to you that it just--
Faggin:
No, it wasn’t obvious. But first of all, as you know, there is never something that is done where you cannot find some prior art, something close. Now, I didn’t know that somebody had built devices like this, but my boss knew. He never told me. He only said to me: “If you use polysilicon, instead of metal, I think you should be able to make a device which is self-aligned. Do it.”
Vardalas:
Why do you think he did that?
Faggin:
Because Bell Lab had made one such device, though it was done differently, as I found out later. What Bell Lab did was a proof of principle that you could make individual self-aligned gate MOS transistors using amorphous silicon. But you could not make integrated circuits with it. The actual idea of making a self-aligned gate came out of work that was done at various companies. All this I found out later. At Hughes, for example, Dr. Bower put an aluminum gate, just like I did here, and then implanted the source and drain regions after he defined the aluminum gate.
His method, however, does not work and was never used by the industry because the aluminum gate cannot withstand the high temperature required to anneal the radiation damage caused by the ion bombardment and to remove the impurities causing unreliable devices. You need a refractory material to make the gate; the Bell Lab guys used amorphous silicon while other researchers were experimenting with molybdenum. Without knowing about this prior work, I invented one method that combined the work of my predecessors and added a number of perfecting steps of my own.
Vardalas:
Why do you think he didn’t tell you? He wanted you to do it a different way?
Faggin:
I guess. I actually don’t know. Maybe he hadn’t read the paper himself. [Laughs]. He may just have heard about it. It didn’t matter; it was probably as well, since I found a better way. For example, to make this process work you had to use vapor-deposited polycrystalline silicon instead of vacuum deposited amorphous silicon, like the Bell Labs researchers did. Amorphous silicon breaks at oxide steps, that’s why their process only produced individual transistors, to avoid this drawback.
Vardalas:
Did you have to work closely with other more production engineers?
Faggin:
Not so much, though I consulted with specialists in different types of processes. For example, I worked with the engineer that had developed a vapor-phase deposition reactor that could deposit different materials in the silicon wafer, including polysilicon.
Vardalas:
How were you greeted with this result?
Faggin:
My direct supervisors were very happy. In fact, they took that technology when they started Intel! That is, Vadasz, Grove and Gordon Moore and of course, Bob Noyce. Noyce and Moore founded Intel after I had developed the first integrated circuit with that technology, the Fairchild 3708. Bob Noyce was the head of Fairchild Semiconductor, Gordon Moore was the head of the Fairchild R&D Lab where I worked, Andy Grove was his assistant and was very familiar with the silicon gate technology and Les Vadasz was my boss and a friend of Andy (both had left Hungary in 1956).
Vardalas:
Because of prior knowledge, this wasn’t patentable, right?
Faggin:
It’s a long story. They didn’t want it to be patented. If it had been patented, it would have been property of Fairchild, not of Intel.
Vardalas:
Ah. I see. So it had nothing to do with prior knowledge?
Faggin:
No. In fact, it’s an interesting story.
Vardalas:
I see. So this is something that stayed in the public domain.
Faggin:
In June, 1968, Gordon Moore urged me to give a paper on silicon gate technology at the International Electron Device Meeting, to be held in Washington D. C. the following October, 1968. This decision was less than a month before he co-founded Intel…
Vardalas:
And Fairchild didn’t have a clue?
Faggin:
Didn’t have a clue.
Vardalas:
I guess intellectual property lawyers weren’t around.
Faggin:
Not exactly, but in those days everybody was taking technology from everybody else. It was a different time, with less attention paid to these aspects of the business.
Impact of silicon gate technology
Faggin:
This technology was absolutely the key to give us the performance, the cost-effectiveness and the reliability that was needed to make semiconductor memories and the microprocessor. It also made possible CCD image sensors and floating gate structures, indispensable to make non-volatile memories, like flash memories, for example. You cannot make non-volatile memories with aluminum gates. All these benefits, in one stroke, removed the fundamental limitations holding back the progress of MOS technology. Within a short time, MOS technology could overtake bipolar technology, which was then by far the dominant technology.
Vardalas:
Because as I understand it, the problem of MOS was the scalability. And that couldn’t be happening unless you had this thing.
Faggin:
That’s right. The silicon gate technology immediately gave us a speed improvement of a factor of three to five. Three to five times more speed while dissipating less power. So the speed-power product was about ten times better than the older technology. Circuit density was also improved by a factor of two, giving us about half the size, therefore half the cost, for the same electronic function. And then we could finally make reliable MOS chips. Reliability was a major problem with MOS before silicon gate, making it a non-starter for many applications. With silicon gate we could then confidently start the process of scaling down the critical dimensions that took us to this day.
Vardalas:
Especially when you’re trying to compete against bipolar.
Faggin:
Yes. If you have threshold voltages that drift, forget it. And most processes with metal gate were tricky and difficult to maintain clean. There are also more subtle reasons why silicon gate technology was so good. For example, junction leakage was dramatically reduced, so you could make better dynamic circuits – like dynamic memories -- and better analog devices than with metal gate, because you could clean up the system. To effectively clean up the system, you have to do it after the entire MOS structure has been sealed, and using temperatures that the aluminum gate cannot sustain. That’s why a refractory gate, like polysilicon, is needed.
Vardalas:
When the phrase Moore’s Law keeps being thrown around. People, in my view, think it’s a law about semiconductors. But it’s a law about MOS. It wouldn’t have worked with bipolar.
Faggin:
Yes. It worked with bipolar, too, early on.
Vardalas:
But the extent that it couldn’t go any further.
Faggin:
Yes. With bipolar, it couldn’t go any further. MOS took over after that.
Vardalas:
And with this kind of thing, there were no limits.
Faggin:
It kept on going, because MOS really scales well, where bipolar doesn’t scale as well because it’s a bulk device, not a surface device. You cannot scale bipolar transistors in the vertical dimension to the same extent you can in the other two dimensions.
Vardalas:
So you left Fairchild, obviously. You were dissatisfied? Your six months were over, and you had to make a decision.
Decision to stay at Fairchild; departure of Fairchild employees for Intel
Faggin:
Yes, but as the end of my six-months stay was approaching, I decided that I wanted to remain in this country. I had already found a job at HP Labs. But then, at about the same time I was having discussions with HP, Fairchild decided to sell their interests in SGS Fairchild -- I don’t know why they made such decision.
At that point, they asked me if I wanted to stay with them, and I accepted the offer. My only reason to go to HP Labs was Fairchild’s difficulty in hiring me, had they still been partners with SGS. But now that they were splitting, this option was open to me, so I decided to stay.
Vardalas:
But eventually, you left. I get the impression you were dissatisfied. What made you decide to stay with Fairchild?
Faggin:
First of all, Noyce and Moore left to start Intel – By the way, it happened exactly on my very first day of employment with Fairchild. A week later Andy Grove left and joined them, then Vadasz, my boss, was gone. Within a short period of time Intel took about 20 Fairchild employees, the best people. Fairchild never recovered from that and over time I became dissatisfied, but not at the beginning.
Vardalas:
I don’t know all the details on the inside, but Fairchild didn’t try to fight this legally? They couldn’t in those days?
Faggin:
I don’t know why Fairchild didn’t respond legally, although it is my impression that the industry was less litigious then. Even worse though, they didn’t even respond competitively. I remember that when Vadasz left, a couple of weeks after Noyce and Moore left, I smelled a rat. I figured that they were going to adopt silicon gate. I immediately told the boss of my boss, Bob Seeds -- because my boss was Vadasz, and they had not yet replaced him. I went to his office and I said: “I suspect that Intel is going to do silicon gate. I know that both Vadasz and Grove thought very highly of this technology and they are both gone.”
First I suspected it, but then, after they hired the technician who was doing the deposition of the silicon for me, I became convinced. That was it! That was the smoking gun, as far as I was concerned. So I went to see Bob Seeds again and I said, “Hey look, they hired this guy. I think they’re going to take our technology!” And he said: “Oh, don’t worry about it. If they do it, we’ll sue them. They cannot do that.” But Fairchild didn’t sue Intel, even after it became a fact they had taken our technology”
Vardalas:
And the rest is history.
Faggin:
The rest is history.
Vardalas:
Were you disappointed that you weren’t asked to join this company? Did they ever come to you directly when this was forming?
Faggin:
No, they didn’t want me at that point, because – speaking of smoking guns – had they had hired me then, their plan would have been too obvious. So they didn’t ask me.
Vardalas:
So you’re saying now that they raided Fairchild fairly extensively after they formed Intel. So you’re trying to imply that it was fewer and fewer people in your area, that the morale had gone down? We’re trying to understand why you left.
Faggin:
Yes, because Fairchild never recovered from the exodus of its best people. Of course, Fairchild was already on its way down, so it would be unfair to say that the company collapsed because they lost those 20 people. But that was the straw that broke the proverbial back, in a way. The departure of those people broke the morale of Fairchild.
Vardalas:
Was Fairchild committed to MOS too?
Faggin:
They were a big player in MOS but, for example, one frustrating thing for me was the delay in the transfer to production of the silicon gate technology. My immediate response to my hunch that Intel was going to adopt the silicon gate technology was to urge Bob Seeds to accelerate its transfer to production and commit to designing new products with it. I even went as far as asking him to send me to Mountain View to set up the technology in the manufacturing line, even if I hated to do it, preferring to continue my work in the Palo Alto R&D group. Mountain View was the location of headquarters and also where product design and manufacturing were done. The R&D was a sort of ivory tower, separate from operations, and had poor communications with the rest of the company.
But nothing happened. Also, surprisingly, the chip designers had not understood the potential of the new technology because the way to design with it was not obvious. They were complaining, saying to me: “Your silicon gate technology is no good because it takes too much area.” And they showed me circuit layouts they had made to prove the point. They were horrible, because they didn’t understand how to use the new technology. They didn’t think in its terms.
Vardalas:
They had to do it the old way.
Faggin:
The old way. They were using the same basic patterns they had for metal gate with the silicon gate. Of course it would come out bigger if you did it that way. So I showed them how to do it, but it still didn’t help. I was frustrated, because I was sure that Intel was going to soon come out with integrated circuits with silicon gate. Seeing the difficulty in getting the silicon gate to production at Fairchild and the inertia of the company, I felt they had lost their sense of urgency and their zest for excellence.
Transition to integrated circuits design
Faggin:
So I started thinking about leaving. I had an unsolicited job offer from National Semiconductor who wanted to hire me to set up the silicon gate technology process for them. But I didn’t want to do that, because I didn’t want to repeat what I already had done; it didn’t make sense to me. At that point in my career, I wanted to design integrated circuits. I didn’t wish to do process technology development anymore. I had done quite a bit of it at SGS and at Fairchild. Before leaving Fairchild, for example, I was developing a silicon gate BiCMOS process technology, combining complementary MOS with bipolar transistors. I had already developed N-channel and CMOS silicon gate processes.
It was amazing, what we were doing in those days. It took many years before silicon gate N-channel and then CMOS became mainstay technologies, yet I had done all that work at Fairchild by 1969.
Vardalas:
Before we go on to when you join Intel, as a general issue, did your heavy experience with process technology have any benefits for you when you started designing circuits?
Faggin:
Absolutely. First of all, the design engineers didn’t know how to use silicon gate. They didn’t know how to optimize circuits and layouts with it. I had to develop all the methodology of how to do random logic with silicon gate from scratch – the way to do it was very different from metal gate. Nobody had done it before. Also, how do you take full advantage of the silicon gate technology itself? There were all kinds of different tricks you had to use, a whole new set of methods, circuits and procedures had to be figured out. That’s what I did with the 4004, laying the foundation for the design of all the first, second and third generation microprocessors.
Vardalas:
I want to get to that, but I also want to get to this whole bag of tricks and the way you fixed circuits when you know how they work a little bit later.
Move to Intel; the 4004 project
Vardalas:
So how did you finally get to Intel? I would imagine, now, they don’t want to see your face. Or did you want to see them? How did you finally get to Intel?
Faggin:
I think the straw, if I remember correctly, was this job offer from National. I did not seek them out, they came after me, but through that process, I became convinced that I had to leave Fairchild. Then, as I said, I wanted to design circuits with silicon gate, because I was totally convinced that it was the best technology -- it was also my baby and I wanted to see it thrive. Since the only other company that had silicon gate, other than Fairchild, was Intel, I called Les Vadasz and I said: “Les, I want to design circuits. I’m tired of working here at Fairchild. Do you have a job for me?’
It took a little while, probably a few months, and then he called me up asking if I could do logic design. I said, “Yes. I designed a computer when I was a kid. I can do it!” [Laughs]. He continued, “I have a project here… I cannot tell you anything about it, but it’s a project for a customer and it involves a fair amount of logic. I would like you to do it.” “Sure, yes I want to do it,” I said.
Vardalas:
Was this the 4004?
Faggin:
Yes. That was the 4004.
Vardalas:
Oh, okay. So that was your first assignment? That was the reason you went to Intel, for that project?
Faggin:
Yes, but I didn’t know what I was getting into. All I knew was that the project was aligned with what I wanted to do, which was to design integrated circuits.
Vardalas:
What did you find when you started? Was the 4004 just from blank pages started when you got there? Was there anything?
Faggin:
It was the specifications, meaning the set of instructions, the block diagram, and the basic architecture of the 4000 family, which was done by Ted Hoff, with some help from Stan Mazor. Ted Hoff also had done some trial circuits for the 4001 and 4002 – not for the 4004 -- that later revealed themselves useless and I discarded. Ted Hoff, did not claim to be a circuit designer.
Vardalas:
So you integrated the logic design and the circuit design.
Faggin:
I had to do the logic design, circuit design, layout and everything else, including completing the architecture that still had some unresolved issues. But it wasn’t routine, because circuits of that complexity had never been done before, particularly with silicon gate. It is curious that at the same time there was another microprocessor project going on at Intel, the 8008.
Vardalas:
At the same time?
Faggin:
Same time. That was a custom project for Datapoint, using Datapoint’s architecture. It was not an architecture done by Ted Hoff or by anybody else at Intel. It was, in fact, very similar to the 4004, because all small computer architectures – if you know enough about computers – are very much alike.
But Hal Feeney, who was hired to do the design, had some trouble getting off the ground. There was no design methodology and he didn’t know enough about silicon gate to figure it out. So the project was mothballed, and it was resumed after the 4004 was completed. I later took over the project, with Hal Feeney working for me. By following the methodology I developed for the 4004, it was relatively easy for Hal to complete it.
Challenges of the 4004 project
Vardalas:
The 4004, you said, “Yes, sure, I can do logic design. I can make a computer. I did it fifteen years ago.” Were there any special challenges you discovered, or dead ends? When you did this, were there some very frustrating moments, or was it one step at a time?
Faggin:
It was very difficult. My frustration, actually, was not in the technology itself, but in the fact that Intel had promised Busicom, the Japanese customer for the 4000 family, that all four chips would be completed six months after I joined the company. There was no way I could possibly make up for lost time, not to mention the fact that the original development schedule – prepared by Vadasz -- was overly optimistic. For example, Vadasz had scheduled the layout time for the 4004 to be about the same as that for the 4001, the ROM chip. He clearly did not appreciate the magnitude of the problem, having no experience with random logic designs. Even without having done before a chip of that complexity myself, I could immediately see that the 4004 would take five to ten times more layout time. So I was late even before I had started!.
Vardalas:
That’s why they hired you [laughs].
Faggin:
Yes. Basically, the project had been sitting there for six months with no work being done. Masatoshi Shima, an engineer of Busicom, arrived at Intel one or two days after I joined the company to check on the progress. He expected to find the logic design of all four chips completed and was very upset when he found that absolutely no progress had been made since his last visit, when the basic architecture was proposed. Nobody had alerted him to the delay and he was mad at me, the project leader, for misdeeds I had not done. He kept on saying, “this is just idea!” and called me a bad manager. I ended up holding the bag.
It took me a while to calm him down. He eventually understood that I could not have done in a couple of days the work of six months and he agreed to help me catch up on a schedule that was irreparably compromised.
Vardalas:
There was never an attempt to try to make the 4004 and the 8008 come together at the same time? They never saw both as similar?
Faggin:
No. They were separate projects.
Vardalas:
They never saw any synergy between them?
Faggin:
Of course they saw synergy between them in the sense that they were two similar CPUs.
Vardalas:
But they don’t have a common methodology?
Faggin:
But you see, my job was not to design one chip only. I had four chips: the 4001, 4002, 4003, and 4004. The 4001 was a 2000-bit ROM plus I/O, and that was a state of the art chip. The 4002 was a 320-bit dynamic RAM. It was a self-sufficient RAM with its own refresh circuitry plus a 4-bit output port. At that same time, Intel was designing the 1103, a state of the art 1000-bit dynamic RAM, but it was only a dynamic RAM; it didn’t have the refresh circuitry and all the additional logic required to be part of the system.
Then there was a shift register, the 4003, the only chip that was simple. It was a 10-bit static shift register with serial-in, serial-out and parallel-out data that was used to facilitate the I/O function. And then there was the CPU, the 4004. So I had to design all those chips. And the right sequence of doing them was actually how I named them: 4001,2,3,4.
Vardalas:
Was that the right sequence, do you think, in retrospect?
Faggin:
Yes, it was the right sequence, because I wanted to learn as I was going through the entire project. That was the step-by-step process I set for myself, because that way I could develop the design methodology a little bit at a time. Since the CPU had ROM, RAM and random logic, it would have been much too soon to start with it. Hal Feeney, though, would have had to do the whole thing at once.
Looking back, it was a daunting job given that there was no methodology to do this kind of project. Compared to Hal, I could ease into the 4004 design having learned from doing the other three chips.
Vardalas:
Did you feel panic?
Faggin:
No, of course not. That’s the power of not knowing any better!
Vardalas:
Were they giving you enough resources to do it?
Faggin:
No. I even had to train my own draftsmen. The first draftsman was a mechanical draftsman out of Lockheed, hired three weeks late and without any chip layout experience, not even experience in printed circuit board layout. All I could rely on was his generic skill with pencil and paper. I had to free-hand sketch the layout of each and every block of the 4001 and he would properly copy it in the composite layout at its assigned place.
Vardalas:
So Intel, then, really didn’t think much of this project?
Faggin:
No.
Vardalas:
They just had to do it because there was a contract?
Faggin:
They had to do it because they had taken a contractual obligation. The attitude of my boss was: “Go do it and leave me alone, I’ve got more important projects to take care of.” You see, Intel’s management had decided to get into the custom business because sales of semiconductor memories were much less than what they expected and they needed to ramp up sales faster than the memory business allowed. Those chips were fillers, and Vadasz’s heart was not in them.”
Vardalas:
So all this legend about this--
Faggin:
That’s bullshit. Complete nonsense.
Vardalas:
Okay. So you answered my question, how did Intel perceive the 4004?
Faggin:
Just a pain in the ass because “Our business is memory.”
Vardalas:
Their core business was memory.
Faggin:
“Our business is memory, and all this other stuff is filler.”
Business and historical significance of the 4004
Vardalas:
In retrospect, and historically, how would you assess the importance of the 4004, of this project, from a technical and business perspective for Intel? Obviously it was crucial in the long term.
Faggin:
Actually, from a business standpoint, it was crucial for them, even in 1971. They actually made a lot of money with the 4000 family that year, because they charged high prices to Busicom, and their volume was pretty high. 1971 was a recession year and a crucial year for Intel because they were still in start-up mode. From a technical perspective, the random logic design methodology I developed was used for the next few generations of microprocessors and microcontrollers and the silicon gate bootstrap load and buried contact I had invented and which I used in the 4004 ended up being used throughout Intel chips. The bootstrap load became obsolete only with the introduction of the depletion load, another of my contributions at Intel.
Vardalas:
And Busicon gave up its rights—in principle, it should have been Busicom property, right?
Faggin:
Well not quite. That’s a misconception.
Vardalas:
What is the truth then?
Faggin:
It is true that Busicom had the exclusive right to the use of the 4004, but they did not have any rights to the 4004 intellectual property. It’s a critical distinction, and some people don’t put it all together. Basically, Intel could have designed another CPU, just a bit different, and it would have been fine. Busicom had no rights either to the 4004 tooling or to make or sell the chip. In fact, they could not even use another manufacturer – other than Intel -- to have the chip made for them. All they had was the exclusive right to the use of the 4000 family, not to make it, have it made or sell it.
The 8080; Intel's status as a microprocessor company
Vardalas:
Was the 8080 the point where Intel saw itself as a microprocessor company?
Faggin:
No, although the 8080 was very important. The 8080 just got Intel to begin to appreciate microprocessors and what they could do for them. No, it was with the 8086 and the success of the personal computer market.
Vardalas:
Okay. Did you have a role in the 8080?
Faggin:
Oh, yes. It was my idea and my architecture. And I pushed very hard for it. Intel’s management took nine months before deciding to proceed and give me the resources to do it. I also directed the whole project.
Vardalas:
From a management perspective, it must have been quite a frustration for you.
Faggin:
Absolutely frustrating, because they didn’t see what was happening and I felt they didn’t care. We almost lost our leadership to Motorola that introduced their 6800 microprocessor about six months after Intel came out with the 8080. If Intel had done what I wanted to do, we would have retained our entire competitive advantage, a full product generation ahead of the competition.
Vardalas:
Was it easy for a company that defined itself as a memory company to change gears and for both management and engineers to become a microprocessor?
Faggin:
It was not easy at all. In fact the company went through a wrenching time in ’85. That’s when they bit the bullet.
Vardalas:
What does that mean, they bit the bullet?
Faggin:
Basically, they had to decide whether they were going to be a memory company or a microprocessor company. I don’t think that was an easy decision?
Vardalas:
Was there a lot of internal fighting over this?
Faggin:
I wasn’t there anymore, so I don’t exactly know. But I read an account of it, saying that it was a very wrenching time. On the other hand, the semiconductor memory business had been decimated by the Japanese competition and it was no longer a profitable business, so the decision to commit to the microprocessor business was a bit like making virtue out of necessity. And the rest is history. This happened way after I left Intel. Ten years after I left. Before that time, Intel was a memory company.
CMOS in supercomputer technology
Vardalas:
Okay. I want to get to Zilog. But, before that, there is something that interests me very much: the relationship between bipolar and CMOS. The reason I say this is, I’m doing my own research on supercomputer industry and the Control Data Corporation and Cray. The dynamics that were going on in this transition period, when you have bipolar and CMOS coexisting, and especially in the days when the company designed large mainframe computers. I know there were intense debates within their own engineering staff about what’s the value of bipolar versus MOS. I know, for example, in my own work looking at Control Data, they saw little use for MOS, mainly because nothing could match the speed of bipolar. Can you recall this debate?
Faggin:
But that’s not entirely true, because for mass storage, MOS dynamic memories became a mainstay of mainframes by the late 1970s.
Vardalas:
But Control Data, didn’t they use CMOS technology? Even Cray computers didn’t start using it until the 1980s.
Faggin:
Yes. You’re talking about supercomputers.
Vardalas:
Yes, I’m talking about supercomputers.
Faggin:
But generic computers, no. For supercomputers, you’re absolutely right.
Vardalas:
They kept postponing.
Faggin:
Absolutely.
Vardalas:
But of course, CMOS saved them all these interconnect problems by making it compact.
Faggin:
I’m not an expert of supercomputer technology, but I would say, looking from the outside in, that the bipolar speed-power product, which is one of the key figures of merit of a semiconductor technology, is substantially worse than CMOS, probably a couple of orders of magnitude worse. It’s a huge, huge difference. Not to mention the large difference in circuit density between bipolar and CMOS.
Vardalas:
People were very cautious. It was not an issue of power. Performance was the issue. That’s all they cared about.
Faggin:
That’s right. But eventually brute force is not the way to go. And if you have something that is substantially better in those metrics, even if all you care is speed, you can actually take a different direction. You don’t get speed directly, but you get it in a different way. How do we make supercomputers now? By combining thousands of microprocessors. Right? In other words the best computer architecture ought to be inspired by the best technology you have available. So the people that insisted on making computers by pushing the limit of the old way of doing things, did not see coming the continuing improvement in CMOS technology that would have allowed them to develop machines in a radically different way.
Vardalas:
I wanted to ask you this, because CMOS was a disruptive technology for the bipolar people in a sense. Now looking from the outside, I’d like to ask you to speculate.
Faggin:
I remember one of the problems I had at Intel was to develop a fast static memory using 5 volt supply. Intel had designed the 2102, the world’s first one-kilobit, 5 volt, N-channel static RAM, and it was very slow -- the access time was 1.5 microsecond. They had a customer, Borroughs, which had ordered a lot of chips with 500 nanoseconds access time – a selection from the normal production – but Intel could not make them. Their yield at 500 nsec was very low, so they couldn’t deliver enough parts to satisfy Borroughs.
Vardalas:
What year would this be about by the way?
Faggin:
This was 1974. Vadasz, in desperation, asked me to redesign the 2102 and wanted me to do it using thinner gate oxide. After I studied the situation, I told Vadasz that the correct approach was to use depletion loads, possibly in combination with thinner oxide. He didn’t want me to do it. I found out later, by the way, that he didn’t want it because he had written a memo to Gordon Moore saying that depletion load was not a good approach, recommending instead to use thinner gate oxide.
Anyway, I ended up in a fight and eventually I told Vadasz: “Look, if you want to do it your way, find somebody else to do it because it won’t work. If you want me to do it, I’ll do it my way.” Vadasz said angrily, “Okay, if that’s what you want, do it!” and he slammed the door and left. Fine, so I did it. Six months later we had the new memory designed with ion-implanted depletion loads and it had 80 nanoseconds access time! A huge improvement.
N-channel MOS with depletion load was the beginning of the end for bipolar memories -- in those days fast static memories used bipolar isoplanar technology, and this technology was their last act before MOS took over. They were memories in the range of twenty to forty nanosecond access times, and here, in one shot, even without CMOS we had closed the gap down to eighty nanoseconds…When I left Intel, the engineer who under my direction did the detailed design of the memory, Dick Pashley, took over in the new direction. Naturally, he is credited by Intel for having done the entire thing right now. It’s normal at Intel. They tried to erase my name from all of my contributions, including the silicon gate technology and the first microprocessor, and attribute them to others. Their official reason was to punish me for having started Zilog, but I believe that erasing the tracks of where their silicon gate technology came from was a strong motive.
Anyhow, it was at that time that MOS memories began to penetrate deeper and deeper into bipolar territory, to make cache and other fast memories. CMOS technology happened six-seven years later, giving us static memories with speed levels that were not much better than N-channel MOS with depletion load, though they had much lower power dissipation. With lower power dissipation per bit, we could then put more bits per chip and accelerate down the scaling potential of MOS.
Vardalas:
Imagine now you’re a circuit logic designer for a high performance mainframe computer and you’re using bipolar, and now you have your whole bag of tricks. You know how to tweak circuits, how to do this, how to get the most performance you can out of this. Was it easy for them to move into a CMOS mentality? Maybe they resisted CMOS because it challenged everything they knew and understood.
Faggin:
There is no question about that. It challenged everything they understood. And their idea was to go from bipolar to gallium arsenide.
Vardalas:
That’s right. I remember Cray was obsessed with gallium.
Faggin:
Yes. That was the path they chose, skipping completely silicon-based CMOS.
Vardalas:
Now was CMOS easy? You started off fairly early, but for somebody who is thirty, forty years old and is a top designer of bipolar high performance computers, would it have been easy for him to make the switch to that?
Faggin:
Not easy.
Vardalas:
That’s what I’m saying, would it have been difficult to think in terms of CMOS and tweaking CMOS?
Faggin:
Not easy, since they didn’t have good CMOS designers and they were buying memories, not making their own, so they had never learned how to design fast CMOS chips. But I don’t think that this was the issue, though. I think the point was the way the architects and the computer designers saw the situation.
Vardalas:
They would design their own Central Processing Units. They designed all processors.
Faggin:
Yes, but they would design it out of building blocks, losing a lot of performance in the connections among chips. They were not single-chip solutions. In a typical computer, the random logic of the CPU was done with bipolar PLDs, programmable logic devices -- maybe not at the Cray level. In fact, I don’t know how the Cray machine was built.
Vardalas:
Cray was very “one of” kind of thing.
Faggin:
Yes. But I don’t know how many custom circuits they actually used for their processor. I would be surprised if they used a lot of custom devices because it would have been very expensive.
Vardalas:
$20 million a computer.
Faggin:
Yes, but it costs you a great deal of development money and time too if you do a lot of customization.
Vardalas:
Historically the companies that developed the transistors were not the same electronic companies that produced vacuum tubes. You also find that the vacuum tube designers had so much invested know-how in the vacuum tube that it was not easy for them to accept the transistor. So I’m wondering if it was the same thing you’re bipolar designer.
Faggin:
Yes, but in this case it was not the designer of the components that could not make the transition. It was the designer of the computer. The computer architects and the overall designers were the ones that had invested heavily in the old technology and were convincing each other that it was the way to go, oblivious to what was going on all around them. It’s also likely that the bipolar chip engineers, not knowing MOS, were also feeding their delusion. It was a sort of inbred society that was getting smaller and smaller and more out of touch with reality. I have noticed that his is the kind of pattern that almost always happens when there are major technological transitions. People who are afraid of change and don’t want to face up to it, begin to isolate themselves, they close themselves down to outside influences, and in their denial their world begins to collapse around them; it implodes.
Vardalas:
I noticed too that enlarged, you know the mainframe computer designers, more and more of the design, competence was leaving them and going to microprocessing companies. So they were being undermined from below.
Faggin:
It’s true. They were eviscerated of all their best people, their young people, because young people see the writing on the wall. Now, if you are an old pro, you have to start all over again, you have to embrace and learn something new and let go of what gives you your sense of worth. It’s hard to start all over again. A young person has much less investment.
Departure from Intel; formation of Zilog
Vardalas:
That’s true. Let me get now to Zilog very quickly. So you left Intel I guess out of frustration, all right?
Faggin:
Yes, essentially yes. Frustration was a major part of it and I had good reasons for feeling frustrated. I felt unappreciated and unsupported, despite my many contributions. To do something good for the company I had to fight for it, as if I were not in the same team. On top of that, I found out that Vadasz had patented one of the ideas I had disclosed to him when we were both at Fairchild. It was the “buried contact,” a way to make direct contact between polysilicon and junctions, without using aluminum. I used that idea, together with another idea of mine, the “silicon gate bootstrap load”, in the 4004 and they were critical to its implementation. That was the last straw.
But it wasn’t just frustration; it was also the challenge of starting my own company.
Vardalas:
When you first had your first business prospectus, when you first presented, what was the core business going to be?
Faggin:
Oh, microprocessors of course.
Vardalas:
And did you intend to take Intel head-on or find some niche you could go after?
Faggin:
Oh no, take Intel on, are you kidding? [laughs] And we almost succeeded.
Vardalas:
Tell me about Zilog.
Faggin:
Well the Z80 was our first product and it became very successful. It took the business away from the 8080. Zilog was winning in the market, but then IBM’s choice to adopt the Intel 8086 reversed the direction. That was the turning point. By the way, the key reason IBM chose Intel was that our sole investor, Exxon Enterprises, had declared war on IBM.
Vardalas:
So your biggest investor was…
Faggin:
Exxon Enterprises, a subsidiary of Exxon, then the largest corporation in the world, was our only investor. It was essentially the owner of Zilog. So that was it.
Vardalas:
So IBM did you in, then.
Faggin:
Yes, because -- from a technology point of view -- we had better products than Intel.
Vardalas:
In those days better in a sense conceptually or quality and reliability?
Faggin:
Not only just conceptually, but also best-of-class reliability, better performance and better price than Intel. But I also made some business mistakes at Zilog, the biggest being to have allowed Exxon Enterprises to become our only investor.
Vardalas:
Now who used the Z80? Was it Commodore? Who was using the Z80?
Faggin:
The Z80 was used by Radio Shack in the TRS-80 personal computer. The TRS-80, at one point, was the largest selling PC in the market, even more popular than the Apple II. We had hundreds of Z80 customers.
Vardalas:
Tandy, the Tandy thing, right?
Faggin:
Right, Tandy was the owner of Radio Shack. There was also Osborne Computers, they used the Z80 in the first portable PC. Commodore had a line of computers as well.
Vardalas:
I think Commodore got on with the 6800, right, with the Motorola.
Faggin:
They used the MOS Technology 6502, which was one generation after the 6800. It was designed by the same engineers that designed the 6800 and had left Motorola to join MOS Technology.
Vardalas:
When did the idea of second sourcing finally end? Or did it ever end?
Faggin:
Obviously Intel did not want to have a second source, but since IBM was such a powerful customer, they had to give IBM the rights to become their own second source. Intel also gave a second source to AMD for the 8086. It was the price to pay to hurt Zilog, since AMD was the official second source of the Z8000, the competitor to the 8086. Later AMD and Intel got into an acrimonious legal battle over second source rights for the x86 products. And Intel has not second-sourced any more follow-on x86 products. Intel was very lucky indeed because with IBM’s decision to have an open architecture, other companies could produce compatible PCs and the entire PC market ended up adopting that architecture. With IBM taken care with a manufacturing license and the smaller competitors having to copy IBM, Intel could get away without a second source.
Vardalas:
Right because IBM had the capacity to second source for itself.
Faggin:
Right. Even if IBM’s competitors had insisted on having a second source, what could they do? They would have had to buy from Intel or get out of the business. So that was the key. Intel could ride on the coattails of the success of IBM’s strategy and get away without a second source, only because IBM -- the only company with the muscle to demand a second source -- had their own semiconductor manufacturing capability and were given the right to be their own second source. A brilliant and lucky solution.
Comparison of engineering and entrepreneurial work
Vardalas:
What gives you most joy, being an entrepreneur or being an engineer?
Faggin:
I think that for me the transition from engineer to entrepreneur has not been without its challenges. My personality was in many ways a good fit, but my education was not. I was much more of an engineer. But I wanted to be an entrepreneur. One reason was to accelerate my personal growth, by creating the conditions where I had to learn how to succeed in a world where machines are not the important thing, but people are.
If you remember, back when we discussed my upbringing, I said that machines were friendlier to me than people because I couldn’t really understand people. People could be mean and deceitful, where machines I could understand and I could make them do what I wanted. I realized that understanding and learning how to succeed in business was not an engineering problem, as you well know, and I wanted to acquire those skills. I have to say that acquiring those skills has been a long process for me, although eventually I’ve been reasonably successful in business. For a long time I definitely got more joy out of my engineering work; only now I am beginning to really enjoy being a CEO. Now I know how to do it; I know the moves, the countermoves; I’ve finally figured out the “machine”!
Vardalas:
I was fixing to say you finally figured out the human machine.
Contributions to the communications field and the computer industry
Vardalas:
In what ways have your particular inventions or breakthroughs contributed the progress of communications? I mean CMOS, I guess, is the basis of communications.
Faggin:
There wouldn’t be a cell phone if you couldn’t have a computer in it. Having developed the microprocessor and having created the silicon gate technology that made the micro and most other telecom chips possible, I believe that I made more than a passing contribution to the progress of communications. However my work was not tied directly to communication, exception made for my second company, Cygnet Technologies, where we developed an intelligent voice and data phone, a companion to the personal computer. Indirectly, the impact has been enormous.
Vardalas:
It’s interesting that how historically how things have changed. There was a point when communications were seen as appliances to computers. Now computers are seen as appliances to communication networks.
Faggin:
In a way, yes. In another way, communications could be seen like a particular I/O function of a computer.
Vardalas:
Except before communications served computers; now computers serve the communications system. Now, your inventions, the work you’ve done, the technical development, has it been used in ways you did not expect? Are you surprised in any of the way it’s turned out from the time when you first initially thought of these things? Did you think it would be a mass phenomenon, the computer and all of this?
Faggin:
In many ways I have been surprised, because the way things happen are most often different from the way we expect them to happen, even when we imagine the very thing that happens. Let me give you an example. When I was designing the 4004, it was clear to me that there would be real computers in the desks of people. No question about it. There were already computers in the desks, the HP 9100 for example, and other scientific calculators. And it was clear that with microprocessors we could make relatively low cost computers for personal use. But I had no idea of the social dimension of the phenomenon, and that the PC would evolve the way it did.
Basically, the birth of a new industry came out of nowhere, out of kids’ enthusiasm. You know, Steve Jobs and Wosniak and Bill Gates and so on. They were kids coming from the outside, pushing against all of the forces of the preexisting order. The milieu that spawned the personal computer was the computer club where kids were playing around, exchanging ideas and dreams. The traditional computer industry was in denial for many years, completely caught by surprise.
Vardalas:
You gave them the tools though to do it.
Faggin:
Of course we gave them the tools. But then the way it happened and the way society adopted it, that was fascinating to see, and I had absolutely no idea that it would happen that way.
Vardalas:
Did you ever imagine it would be so ubiquitous?
Faggin:
No, not at that the point. My notion was like a theoretical idea, like it would be good to have a computer and you could do many things with it, and if they were cheap enough there would be lots of them, but that was it. That was the end of it. I wasn’t passionate about it. In practice, I couldn’t conceive how we could go about making it happen in the world. I had no idea that the computer could become a multimedia center, the hub of the home entertainment, for example. Not to mention Internet. I definitely did not predict Internet, and again, not so much the technology of Internet, but the sociology of Internet, and the way you use it and the things that you can do with it. This is what is fascinating to me.
Vardalas:
I asked Kleinrock, he was one of the first to develop it, and he never imagined the Internet would be a mass consumer phenomenon. He always thought it would be university labs talking to each other, and that’s it.
Faggin:
Yes.
- Audio File
- MP3 Audio
(442_-_faggin_-_clip_2.mp3)
You think of it as a way of sending and receiving data faster, right? Okay, fine, that’s good. But today if I want some information about anything, I click on Google and I get what I need… This stuff is mind-boggling. I certainly had no idea that you could one day do this. Part of it is that a technologist thinks more about the infrastructure, the enabling aspect, and less about the applications, the use aspect. But part of it is also that if I can envision a sufficient number of applications about a new idea to consider worth doing it, I am not going to spend a lot of time thinking about what else can be done with it. For example, right now I see an incredible potential in embedded cameras, tiny little imaging systems that can be embedded in a variety of products for a variety of applications. I can see a few clear cut applications, but then in twenty years I’m sure I will be surprised by how many applications I couldn’t even envision today. I expect to witness in this field the same thing that happened with the microcontroller and the microprocessor.The microprocessor that powers your PC is similar to the image sensor that replaces the photographic film in a camera. The user interacts directly with the PC and the camera. However, a microcontroller is an entire computer on a chip. It incorporates a microprocessor, just like an embedded camera is an entire imaging system incorporating an image sensor. Both are used for all sorts of applications where the user doesn’t even know there is a computer or a camera in it, and doesn’t interact directly with them.
Today, for every microprocessor sold there are twenty or more microcontrollers sold and I expect that the same phenomenon will happen with image sensors; for every digital camera sold, there will be ten or twenty embedded imaging systems sold, for applications that have nothing to do with taking pretty pictures. Embedded imaging systems, like microcontrollers, will find applications everywhere, in cars, airplanes, homes, factories, toys, you name it. They will be the eyes of our machines. They are not cameras, in the sense that their main purpose is not to take pictures to be looked at by a human being or to be saved, or anything like that. The real-time images will mostly be used as part of a control loop of some automatic system, or part of something else. So this is what I now see as the main direction of my new company, Foveon: to develop and market embedded imaging systems for all major applications.
Vardalas:
So it’s part of a feedback group that’s part of some kind of control system.
Faggin:
Yes. But if you were to ask me ten years from now what happened, I’d probably say that in 2004 I could only envision ten applications and that now there are more than 1,000, many of which I could not have possibly imagined. Just like what happened with the microcontroller. And that is fine with me because I see myself more as someone who creates enabling technology than a person that figures out how to use existing technology.
Vardalas:
Thank you for the interview.
