The rays of the rising sun have barely reached the foothills of Silicon Valley, but Marcian E. (Ted) Hoff Jr. is already up to his elbows in electronic parts, digging through stacks of dusty circuit boards. This is the monthly flea market at Foothill College, and he rarely misses it.
Ted Hoff is part of electronics industry legend. While a research manager at Intel Corp., then based in Mountain View, he realized that silicon technology had advanced to the point that, with careful engineering, a complete central processor could fit on a chip. Teaming up with Stanley Mazor and Federico Faggin, he created the first commercial microprocessor, the Intel 4004.
This article was first published as “Marcian E Hoff.” It appeared in the February 1994 issue of IEEE Spectrum. A PDF version is available on IEEE Xplore. The photographs appeared in the original print version.
But for Hoff, the microprocessor was merely one blip among many along the tracing of his long fascination with electronics. His passion for the field led him from New York City’s used electronics stores to elite university laboratories, through the intense early years of the microprocessor revolution and the tumult of the video game industry, and ultimately to his job today: high-tech private eye.
Fairly early in his childhood Hoff figured out that the best way to feel less like a kid—and a little more powerful—was to understand how things work. He started his explorations with chemistry. By the age of 12 he had moved on to electronics, building things with parts ordered from an Allied Radio Catalog, a shortwave radio kit, and surplus relays and motors salvaged from the garbage at his father’s employer, General Railway Signal Co., in Rochester, NY. Then in high school, working mostly with secondhand components, he built an oscilloscope, an achievement he parlayed into a technician’s job at General Railway Signal.
Hoff returned to that job during breaks from his undergraduate studies at Rensselaer Polytechnic Institute, Troy, N.Y. Several summers began with Hoff entering the General Railway laboratory to find the researchers’ two best oscilloscopes broken. He would repair the state-of-the-art Tektronix 545s, then move on to more interesting stuff, like inventing an audio frequency railroadtrain tracking circuit and a lightning protection unit that gave him two patents before he was out of his teens.
The best thing about the job, Hoff recalled, was the access it gave him to components that were beyond the budgets of most engineering students in the l950s—transistors, for instance, and even the just-introduced power transistor. He did an undergraduate thesis on transistors used as switches, and the cash prize he won for it quickly went for a Heathkit scope of his own.
Early Neural Networks
Hoff liked the engineering courses at Rensselaer, but not the narrow focus of the college itself. He wanted to broaden his perspective, both intellectually and geographically (he had never been more than a few miles west of Niagara Falls), so chose California’s Stanford University for graduate school. While working toward his Ph.D. there, he did research in adaptive systems (which today are called neural networks) and, with his thesis advisor Bernard Widrow, racked up two more patents.
“He had a toy train moving back and forth under computer control, balancing a broom stick. I saw him as a kooky inventor, a mad scientist.”
His Intel colleague Mazor, now training manager at Synopsys Inc., Mountain View, Calif., recalled meeting Hoff in his Stanford laboratory.
“He had a toy train moving back and forth under computer control, balancing a broomstick,” Mazor said. “I saw him as a kooky inventor, a mad scientist.”
After getting his degree, Hoff stayed at Stanford for six more years as a postdoctoral researcher, continuing the work on neural networks. At first, his group made the networks trainable by using a device whose resistance changed with the amount and direction of current applied. It consisted of a pencil lead and a piece of copper wire sitting in a copper sulfate and sulfuric acid solution, and they called it a memistor.
“One result of all our work on microprocessors that has always pleased me is that we got computers away from those [computer center] people.”
The group soon acquired an IBM 1620 computer, and Hoff had his first experience in programming—and in bucking the system. He had to deal with officials at the campus computer center who thought all computers should be in one place, run by specialists who handled the boxes of punched cards delivered by researchers. The idea that a researcher should program computer systems interactively was anathema to them.
Ted Hoff: Vital Stats
Marcian E. (Ted) Hoff Jr.
Date of birth
Oct. 28, 1937
Wife, Judy; three daughters, Carolyn, Lisa, and Jill
BS, 1958, Rensselaer Polytechnic Institute, Troy, N.Y.; MS, 1959, Ph.D., 1962, Stanford University, California, all in electrical engineering
First electronics job
Technician, General Railway Signal Co., Rochester, N.Y.
Biggest surprise in career
Media hysteria over the microprocessor
Books recently read
Introduction to Nuclear Reactor Theory by John R. Lamarsh; A Compiler Generator by William M. McKeeman, James J. Horning, and David B. Wortman
People most respected
Intel Corp. founders Robert Noyce and Gordon Moore, Intel chief executive officer Andrew Grove
Postrio and Bella Voce in San Francisco, Beausejour in Los Altos, Calif.
2001, Dr. Strangelove
“If it works, it’s aesthetic”
Playing with electronics; attending operas and concerts; going to the theater, body surfing in Hawaii; walking his Alaskan malamutes
“The best motivation is self-motivation”
IEEE, Sigma Xi
Stuart Balantine Medal of the Franklin Institute, IEEE Cledo Brunetti Award, IEEE Centennial Medal, IEEE Fellow
“One result of all our work on microprocessors that has always pleased me,” Hoff told IEEE Spectrum, “is that we got computers away from those people.”
By 1968 student hostility to the government over the Vietnam War was growing and life for researchers on campus who, like Hoff, relied on government funding was looking as if it might get uncomfortable. Hoff had already been contemplating the possibilities of industrial jobs when he received a telephone call from Robert Noyce, who told him he was starting a new company, Intel Corp., and had heard Hoff might be interested in a job. He asked Hoff where the semiconductor integrated circuit business would find its next growth area. “Memories,” Hoff replied.
That was the answer Noyce had in mind (Intel was launched as a memory manufacturer), and that year he hired Hoff as a member of the technical staff, Intel’s 12th employee. Working on memory technology, Hoff soon received a patent for a cell for use in MOS random-access integrated circuit memory. Moving on to become manager of applications research, he had the first customer contact of his career.
“Engineering people tend to have a very haughty attitude toward marketing, but I discovered you learn a tremendous amount if you keep your eyes and ears open in the field.”
“Engineering people tend to have a very haughty attitude toward marketing,” Hoff said, “but I discovered you learn a tremendous amount if you keep your eyes and ears open in the field. Trying to understand what problems people are trying to solve is very helpful. People back in the lab who don’t have that contact are working at a disadvantage.”
From 12 Chips to One Microprocessor
One group of customers with whom Hoff made contact were from Busicom Corp., Tokyo. Busicom had hired Intel to develop a set of custom chips for a low-cost calculator and had sent three engineers to Santa Clara to work on the chip designs. Hoff was assigned to look after them, getting them pencils and paper, showing them where the lunchroom was—nothing technical.
But the technical part of Hoff’s mind has no off-switch, and he quickly concluded that the engineers were going in the wrong direction. Twelve chips, each with more than 3000 transistors and 36 leads, were to handle different elements of the calculator logic and controls, and he surmised the packaging alone would cost more than the targeted retail price of the calculator. Hoff was struck by the complexity of this tiny calculator, compared with the simplicity of the PDP-8 minicomputer he was currently using in another project, and he concluded that a simple computer that could handle the functions of a calculator could be designed with about 1900 transistors. Given Intel’s advanced MOS process, all these, he felt, could fit on a single chip.
The Busicom engineers had no interest in dumping their design in favor of Hoff’s unproved proposal. But Hoff, with Noyce’s blessing, started working on the project. Soon Mazor, then a research engineer at Intel, joined him, and the two pursued Hoff’s ideas, developing a simple instruction set that could be implemented with about 2000 transistors. They showed that the one set of instructions could handle decimal addition, scan a keyboard, maintain a display, and perform other functions that were allocated to separate chips in the Busicom design.
In October 1969, Hoff, Mazor, and the three Japanese engineers met with Busicom management, visiting from Japan, and described their divergent approaches. Busicom’s managers chose Hoff’s approach, partly, Hoff said, because they understood that the chip could have varied applications beyond that of a calculator. The project was given the internal moniker “4004.”
Federico Faggin, now president and chief executive officer of Synaptics Inc., San Jose, Calif., was assigned to design the chip, and in nine months came up with working prototypes of a 4-bit, 2300-transistor “microprogrammable computer on a chip.” Busicom received its first shipment of the devices in February 1971.
Faggin recalled that when he began implementing the microprocessor, Hoff seemed to have lost interest in the project, and rarely interacted with him. Hoff was already working on his next project, the preliminary design of an 8-bit microprogrammable computer for Computer Terminals Corp., San Antonio, Texas, which, architected by Computer Terminals, was named the 8008. Hoff always “had to do very cutting-edge work,” Faggin told Spectrum. “I could see a tension in him to always be at the forefront of what was happening.”
In those early Intel days, Mazor recalled that Hoff had a number of ideas for projects, many of which, though not commercially successful, proved prescient: a RAM chip that would act like a digital camera and capture an image in memory, a video game with moving spaceships, a device for programming erasable programmable ROMs, and computer-aided design tools intended for logic simulation.
The Intel marketing department they estimated that sales [of microprocessors] might total only 2000 chips a year.
Meanwhile, the microprocessor revolution was gearing up, albeit slowly. Hoff joined Faggin as a microprocessor evangelist, trying to convince people that general-purpose one chip computers made sense. Hoff said his toughest sell was to the Intel marketing department.
“They were rather hostile to the idea,” he recalled, for several reasons. First, they felt that all the chips Intel could make would go for several years to one company, so there was little point in marketing them to others. Second, they told Hoff, ‘‘We have diode salesman out there struggling like crazy to sell memories, and you want them to sell computers? You’re crazy.” And finally, they estimated that sales might total only 2000 chips a year.
But word went out. In May 1971 an article in Datamation magazine mentioned the product, and the following November Intel produced its first ad for the 4004 CPU and placed it in Electronic News. By 1972 stories about the miracle of what began being called the microprocessor started appearing regularly in the press, and Intel’s competitors followed its lead by launching microprocessor products of their own.
Hoff never even considered patenting the microprocessor. To him the invention seemed to be obvious.
One step Hoff did not take at that time was apply for a patent, even though he had already successfully patented several inventions. (Later, with Mazor and Faggin he filed for and was granted a patent for a “memory system for a multi-chip digital computer.”)
Looking back, Hoff recalled that he never even considered patenting the microprocessor in those days. To him the invention seemed to be obvious, and obviousness was considered grounds for rejecting a patent application (though, Hoff said bitterly, the patent office currently seems to ignore that rule). It was obvious to Hoff that if in one year a computer could be built with 1000 circuits on100 chips, and if in the following year those 1000 circuits could be put onto10 chips, eventually those 1000 circuits could be con structed on one chip.
Instead of patenting, Hoff in March 1970 published an article in the proceedings of the 1970 IEEE International Convention that stated: “An entirely new approach to design of very small computers is made possible by the vast circuit complexity possible with MOS technology. With from 1000 to 6000 MOS devices per chip, an entire central processor may be fabricated on a single chip.”
But in December 1970, an independent inventor outside the cliquish semiconductor industry, Gilbert Hyatt, filed for a patent on a processor and mentioned that it was to be made on a single chip. In 1990, after numerous appeals and extensions, Hyatt was granted that patent and began collecting royalties from many microprocessor manufacturers. Currently, though history traces today’s microprocessor back to Hoff, Mazor, and Faggin, the legal rights to the invention belong to Hyatt.
The Invention of the Codec
While the microprocessor has proved to be his most celebrated achievement, Hoff does not view it as his biggest technical breakthrough. That designation he reserves for the single-chip analog-to-digital/ digital-to-analog coder/decoder (codec).
“Now that work was an exciting technical challenge,” Hoff recollected with some glee, “because there were so many who said it couldn’t be done.”
The project was kicked off by Noyce, who spotted the telephone industry as ripe for new technology, and urged Hoff to find an important product for that market. Studying telephone communications, Hoff and several other researchers saw that digitized voice transmission, then being used between central offices, depended on the use of complex expensive codecs that tied into electromechanical switches.
”We thought,” Hoff told Spectrum, “we could integrate this, the analog-to-digital conversion, on a chip, and then use these circuits as the basis for switching.”
Besides reducing the cost of the systems to the telephone company, such chips would enable companies to build small branch exchanges that handled switching electronically.
Hoff and his group developed a multiplexed approach to conversion in which a single converter is shared by the transmit and receive channels. They also established a number of other techniques for conversion and decoding that Hoff saw as not being obvious and for which he received patents.
With that project’s completion in 1980, after six years of effort, and its transfer to Intel’s manufacturing facility in Chandler, Ariz., Hoff became an Intel Fellow, free to pursue whatever technology interested him. What interested him was returning to his work on adaptive structures, combining the concepts he had wrestled with at Stanford with the power of the microprocessor in the service of speech recognition. After a year he built a recognition system that Intel marketed for several years.
A prime customer for the system was the automotive industry. Its inspectors used the systems to help them check out a car as it finally left the assembly line. When an inspector noted out loud various problems that needed fixing, the system would prompt him for further information, and log his responses in a computer.
From Intel to Atari
Though his position as an Intel Fellow gave Hoff a fair amount of freedom, he found himself getting bored. Intel’s success in microprocessors by 1983 had turned it into a chip supplier, and other companies were designing the chips into systems.
“I had always been more interested in systems than in chips,” Hoff said, “and I had been at Intel for 14 years, at a time when the average stay at a company in Silicon Valley was three years. I was overdue for a move.”
Again, Hoff had not gone beyond thinking about leaving Intel when a new job came to him. Atari Inc., Sunnyvale, Calif., then a booming video game company owned by Warner Communications Inc. and a major user of microprocessors, was looking for a vice president of corporate technology. In February 1983, after discussing the scope of the ideas that Atari researchers were pursuing, Hoff latched onto the opportunity.
Intel from the start had a structured, highly controlled culture. At Atari, chaos reigned.
Intel from the start had a structured, highly controlled culture. At Atari, chaos reigned. Under Hoff were research laboratories in Sunnyvale, Los Angeles, and Grass Valley, Calif.; Cambridge, Mass.; and New York City. Researchers were working on picture telephones, electronic aids for joggers, computer controls that gave tactile feedback, graphical environments akin to today’s virtual reality, digital sound synthesis, advanced personal computers, and software distribution via FM sidebands.
But Hoff had barely had time to learn about all the research projects under way before the video game business took a well-publicized plunge. Without solid internal controls, Atari was unable to determine how well its games were selling at the retail point, and distributors were returning hundreds of thousands of cartridges and game machines. Hoff began receiving orders for staff cuts monthly.
“It would have been one thing if I had known I had to cut back to, say, one-quarter the size of my group,” he told Spectrum. “But when every month you find you have to cut another chunk, morale really drops.”
In July 1984, while Hoff was at his 30th high school reunion, Warner sold Atari to Jack Tramiel. Hoff then had to choose between convincing Tramiel that he could play a role in a narrowly focused company uninterested in funding futuristic research, and allowing Warner to buy out his contract. He chose the latter.
Looking back, most of the people who were at Atari in those days now view them darkly. But Hoff recalls his year there as an enjoyable and ultimately useful experience. “Maybe I look at it more positively than I should,” he said, “but it turned out to be a good transition for me, and the life I have now is a very nice one.”
“Whenever you are working on one problem, there is always another problem over here that seems more interesting.”
He now spends half his time as a consultant and half pursuing technical projects of his own devising—a readout device for machine tools, various types of frame grabbers, pattern recognition, and techniques for analog-to-digital conversion. This variegated schedule is perfect for him. He has always felt himself to be a generalist, and has had trouble focusing on just one technology.
“It’s easy for me to get distracted,” he said. “Whenever you are working on one problem, there is always another problem over here that seems more interesting. But now it is more likely that my own projects get delayed, rather than things critical to other people and their employment.”
Faggin for one is not surprised that such independent work appeals to Hoff. “He never was the gregarious type,” Faggin said. “He liked introverted work, the thinking, the figuring out of new things. That is what he is good at. I always was impressed how he was able to visualize an architecture for a new IC, practically on the spot.”
“He comes up with idea after idea, situation after situation. I think if he wanted to, Ted could sit down and crank out a patent a month.”
Said Gary Summers, president and chief executive officer of Teklicon Inc., Mountain View, the consulting firm that employs Hoff today: “He comes up with idea after idea, situation after situation. I think if he wanted to, Ted could sit down and crank out a patent a month.”
“There is no doubt in my mind that he is a genius,” Mazor stated. Summers readily concurred.
Hoff’s first project after Atari was a voicecontrolled music synthesizer, which gave off the sound of a selected instrument when someone sang into it. Hoff’s biggest contribution to the project was a system that ensured that the emerging notes would be in tune, or at least harmonically complement the tune, even when the singer strayed off key. He scored another patent for this system, and the gadget was sold briefly through the Sharper Image catalog, but never became a big success.
Hoff still contributes occasionally to product designs. At Teklicon, however, where he is vice president and chief technical officer, most of his consulting is done for lawyers. Hoff has a unique combination of long experience with electronic design and long-standing pack rat habits. His home workshop contains about eight personal computers of different makes and vintages, five oscilloscopes, including a vintage Tektronix 545 scope, 15000 ICs inventoried and filed, and shelves loaded with IC data books dating right back to the 1960s.
“If my washing machine breaks down, I call the repairman. Most clever engineers would buy the replacement gear and install it. Ted is capable of analyzing the reason the gear failed in the first place, redesigning a better gear from basic principles, carving it out of wood, casting it at his home, and dynamically balancing it on his lathe before installing it.”
When a lawyer shows him a patent disclosure, even one decades old, he can determine whether or not it could then have been “reduced to practice” and whether it provided sufficient information to allow “one of ordinary skill in the art” to practice the invention. Then he can build a model proving his conclusion, using vintage components from his collection, and demonstrate the model in court as an expert witness. This model-building can get very basic. On Spectrum’s visit, Rochelle salt crystals that Hoff attempted to grow for a recent court demonstration littered his workshop floor, next to metal-working equipment that he uses to build cases for his models.
Hoff sees this ability to get down to basics as one of his strengths. “I relate things to fundamental principles,” he said. “People who don’t question the assumptions made going into a problem often end up solving the wrong problem.”
Mazor said, “If my washing machine breaks down, I call the repairman. Most clever engineers would buy the replacement gear and install it. Ted is capable of analyzing the reason the gear failed in the first place, redesigning a better gear from basic principles, carving it out of wood, casting it at his home, and dynamically balancing it on his lathe before installing it.”
Doing legal detective work appeals to Hoff for another reason: it gives him an excuse to hunt for interesting “antique” components at flea markets and electronics stores.
Hoff cannot discuss the specifics of patent cases he has been involved with. Several recently were in the video game area; others have involved various IC companies. In a number of cases, Hoff was confident that his side was right, and his side still lost, so he felt little surprise when the microprocessor patent was granted to Hyatt. (After the award was made, though, he did sit down with Hyatt’s patent application and attempted to design a working microprocessor based on Hyatt’s disclosures. He found several incongruities—like a clock rate only suited to bipolar technology with logic that could only be rendered in MOS technology, and logic that required far too many transistors to put on a chip, proving in his mind that the award was incorrect.)
Seeing someone else get credit for the microprocessor, particularly in recent media reports, “is irritating,” Hoff told Spectrum, “but I’m not going to let it bother me, because I know what I did, I know what all the other people on our project did, and I know what kind of company Intel is. And I know that I was where the action was.”
Editor’s note: Hoff retired from Teklicon in 2007. He currently serves as a judge for the Collegiate Inventors Competition, held annually by the National Inventors Hall of Fame. These days, his main technical interests surround energy, water, and climate change.