Skip to content

Federico Faggin and the Microprocessor

Abstract

Federico Faggin turned the specification of the world’s first commercially available single-chip microprocessor into working silicon in under a year of work, much of it at night, at Intel in Mountain View in 1970 and 1971. The Intel 4004 (2,300 transistors on a die of 3 by 4 millimetres) was laid out by him, in a silicon-gate process he had developed at Fairchild. For decades the public credit went mostly to Ted Hoff. Faggin then founded Zilog, led the Z80, co-founded the touchpad maker Synaptics, and in his seventies turned to the philosophy of consciousness.

Silicon Gate Technology and Fairchild

Federico Faggin was born on December 1, 1941, in Vicenza, a small city in northern Italy known more for Palladian architecture than electronics. He showed an early aptitude for physics and mathematics, and enrolled at the University of Padova (one of the oldest universities in the world) where he completed a degree in physics in 1965.

He joined SGS-Fairchild, the Italian partner of Fairchild Semiconductor, in Agrate Brianza, where he worked on metal-oxide-semiconductor (MOS) process technology. The MOS transistor was emerging as the key device for dense integrated circuits, but it had manufacturing problems. The standard approach used aluminum as the gate material (the electrode that controlled the transistor’s switching behavior) and this metal gate had to be deposited in a separate step from the transistor’s source and drain regions, making it difficult to align precisely. Misalignment increased the transistor’s size and introduced unwanted parasitic capacitance.

The fix was to replace the aluminum gate with polycrystalline silicon, which Robert Kerwin, Donald Klein and John Sarace at Bell Labs had done experimentally in 1967. A silicon gate could be fabricated first, then used as a self-aligned mask for the source and drain diffusions, so the same lithography step that defined the gate also defined where the current-carrying regions would be. This eliminated the alignment problem, making transistors smaller, faster, and more consistent. The technique became known as silicon gate technology (SGT).

In 1968, Faggin was sent to Fairchild’s research laboratory in Palo Alto, where he led the project, with Tom Klein, that turned silicon gates into a manufacturable process, and designed the first commercial silicon-gate integrated circuit: the Fairchild 3708, an analog multiplexer. Intel, founded that year by Fairchild defectors, made the silicon-gate process its main technology for memories, and within five years it was the industry standard for new MOS chips.

The 4004: Architecture Is Not a Microprocessor

When Faggin joined Intel in April 1970, he walked into a situation that tested his abilities immediately. In 1969 Intel had taken a contract from Busicom, a Japanese calculator company, to produce a set of custom logic chips (accounts range from eight to twelve) for a printing desktop calculator, the 141-PF. Intel engineer Marcian “Ted” Hoff, with Stanley Mazor, proposed a simpler solution: four more general chips, one of which (the CPU chip) would be programmable. A programmable processor could be instructed to perform any function, making it far more flexible than a chip hardwired to do one thing.

But Hoff and Mazor were not chip designers, and by early 1970 the project had stalled: nobody at Intel had turned the specification into logic and layout. When Busicom’s engineer Masatoshi Shima arrived in April 1970 to check the logic design, he found almost nothing done. The architecture specified what the processor should do; it said almost nothing about how to fit a complete CPU into the physical silicon area available, using a manufacturing process that had never been applied to a device of this complexity.

Faggin took the project over and immediately understood the gap. Architecture is a specification. A microprocessor is a physical artifact, transistors placed precisely on silicon, connected by metal pathways that have to fit within the heat budget, power budget, and area budget of the die. Getting from specification to silicon requires a kind of intimate knowledge of the manufacturing process (what you can and cannot ask the technology to do) that Faggin possessed and the rest of the team did not.

Info

Faggin later described the process of chip design in 1970 as entirely manual. There were no computer-aided design tools. The circuit schematic was drawn on paper. The physical layout (the precise placement of every transistor and every metal wire on the silicon surface) was drawn by hand on large sheets of rubylith film, which were then photographically reduced to create the masks used in manufacturing. A single engineer who understood both the circuit and the manufacturing process could design a chip no team of engineers who understood only one of these things could match.

Over the following months Faggin, with Shima working beside him, designed the circuits for all four chips in the MCS-4 chipset: the 4001 ROM, 4002 RAM, 4003 shift register, and 4004 CPU. He created the physical layout of each chip. He applied silicon gate technology to achieve the transistor density the design required. He built prototype circuits, ran silicon through the fabrication process, tested the results, identified failures, corrected the design, and ran more silicon. He did much of this work at night and on weekends. The design was finished at the end of 1970, and Busicom had a working chipset in March 1971. Faggin etched his initials, “F.F.”, on the edge of the 4004 die.

The 4004 (the CPU chip) contained 2,300 transistors on a die measuring 3 millimeters by 4 millimeters. It had a 4-bit data path and a 12-bit address bus. At its rated clock speed of 740 kHz, it could execute roughly 92,000 simple instructions per second. By the standards of 1971 minicomputers, its computing power was modest. But it was a complete CPU, including arithmetic logic unit, registers, and control logic, on a single piece of silicon.

Before it could sell the chip to anyone else, Intel had to buy back Busicom’s exclusive rights in 1971, in exchange for lower chip prices to Busicom. Intel announced the 4004 publicly with a two-page advertisement in the November 15, 1971 issue of Electronic News, headlined “Announcing a new era of integrated electronics.”

The Credit Dispute

In the press and in Intel’s later publicity, the microprocessor became Hoff’s invention. Faggin, who had designed the circuits and layout of the chip in a process he had developed, appeared only incidentally, and this version persisted in popular accounts for decades.

Faggin objected, consistently and publicly. His argument was not that Hoff’s architectural contribution was insignificant (it was not) but that architecture without implementation is not an engineering product. Hoff had conceived what the microprocessor should do; Faggin had made it real, through twelve months of design work that required both deep knowledge of silicon gate technology and the ability to apply it to a circuit of unprecedented complexity. Without Faggin, the 4004 would have remained an idea.

Info

The credit dispute around the 4004 was complicated by Masatoshi Shima, a Busicom engineer who worked with Faggin during the design, contributed to the circuit’s debug phase, and has his own credible claim to portions of the design work. A fourth contributor, Stanley Mazor, co-developed the instruction set with Hoff. The four told the story jointly in IEEE Micro in 1996, and the four shared the Kyoto Prize in 1997. In 2010 President Obama presented Faggin, Hoff and Mazor with the National Medal of Technology and Innovation for the microprocessor.

The 8008, Intel’s Growing Indifference, and the Exit

Faggin continued at Intel after the 4004 and led the Intel 8008 (1972), the first 8-bit microprocessor, whose detailed design was done by Hal Feeney under his supervision. He then led the 8080 project, with Shima, now at Intel, as its designer. The 8008 originated from a separate project: CTC, a Texas company, had contracted Intel to develop a terminal processor. The contract was eventually cancelled by CTC, but Intel retained the rights to market the chip. The 8008 ran the first hobbyist microcomputer kits, such as the Mark-8 (1974), and its instruction set architecture directly influenced the 8080 and 8085 that followed.

Faggin’s frustration with Intel grew through 1973 and 1974. He felt that Intel’s management (focused primarily on the memory business, which was generating most of the company’s revenue) undervalued the microprocessor division and its engineers. He also remained aggrieved by the credit situation around the 4004. In late 1974, he left Intel.

Zilog and the Z80

Faggin’s last day at Intel was 31 October 1974, and in December he co-founded Zilog in Los Altos with Ralph Ungermann; Masatoshi Shima, who had worked on the 4004, joined in April 1975. Exxon Enterprises put in half a million dollars in June 1975, when the team numbered eleven. Faggin first planned a microcontroller, then switched to an improved 8080, at first called Super-80. The first chips came from an outside foundry in early 1976, a two-page advertisement in May 1976 compared the Z80 with the 8080 under the headline “The Battle of the 80s”, and it went on sale in July 1976. Zilog had its own fab from January 1977 and more than a thousand employees by December 1978.

The Z80 was designed to be fully software-compatible with Intel’s 8080 (programs written for the 8080 would run on the Z80 without modification) while adding substantially more capability. Faggin defined the Z80’s architecture and Shima did its logic design. It had two complete sets of registers, allowing operating systems and interrupt handlers to switch register contexts without saving and restoring registers in memory. He added an extended instruction set, block move and search instructions for memory operations, bit manipulation instructions, relative branch instructions that reduced program size. He included built-in dynamic RAM refresh circuitry, eliminating external refresh logic that designers had previously needed to add.

The Z80 entered a market where the 8080 was established but expensive. Zilog priced the Z80 competitively and delivered it in a 40-pin DIP package compatible with 8080 systems. It was faster, had more features, and cost less.

The result was one of the most successful microprocessors ever made. The Z80 powered the TRS-80 Model I (Radio Shack, 1977), the Sinclair ZX80 and ZX81 (1980, 1981), the ZX Spectrum (1982), the machine that introduced personal computing to an entire generation of British users, the Osborne 1 portable computer, and countless CP/M-based business computers. The Z80 was the dominant processor of the CP/M ecosystem, the leading personal computing platform before the IBM PC. The Pac-Man arcade machine used a Z80. So did the MSX home computer standard, adopted across Asia and Europe. Millions of units were sold across thousands of applications. East Germany copied it without a licence as the U880, the standard processor of its computers.

The Z80 enjoyed one of the longest production runs of any microprocessor, roughly 48 years. Zilog finally discontinued the standalone Z84C00 CPU family in 2024 (last-time-buy date June 14, 2024), though the related eZ80 embedded core lives on and the Z80 architecture remains widely emulated and cloned.

The 16-bit Z8000, introduced in early 1979, came after Intel’s 8086 (June 1978) and was followed by Motorola’s 68000 that September; it never matched the Z80. Faggin left in 1980, the year Exxon took full ownership. Zilog’s employees bought it back in 1989, it went public in 1991, filed for bankruptcy in 2001 and recovered, and was bought by IXYS in 2009, which Littelfuse acquired in 2017.

Synaptics and the Touchpad

After Zilog, Faggin went through several ventures before co-founding Synaptics in 1986 with Carver Mead of Caltech. Synaptics was initially focused on analog VLSI hardware for neural-network-inspired computation, building silicon circuits that performed pattern recognition by mimicking aspects of neural processing.

The neural network approach led to a more immediately commercial application: capacitive touch sensing. Synaptics developed the TouchPad, a flat, touch-sensitive surface that could track the position of a fingertip through the capacitive coupling between the finger and a grid of sensors beneath the surface. The device required no mechanical parts, no moving surfaces, no lubricant. It was thin, durable, and accurate.

The Synaptics TouchPad became the standard laptop pointing device through the 1990s and 2000s. Virtually every notebook computer sold between 1994 and the present has used either a Synaptics TouchPad or a competing device built on similar principles. Synaptics later supplied touch controllers for phones as well; Apple’s iPhone used its own multi-touch technology, from the company FingerWorks it had bought in 2005.

Consciousness and Later Life

In his seventies, Faggin turned to questions that might have seemed surprising from a microprocessor engineer: the nature of consciousness, the relationship between mind and matter, the limits of computational models of experience. He published a book, Silicon: From the Invention of the Microprocessor to the New Science of Consciousness (2021), in which he argued that the computational model of mind (the idea that consciousness is a form of information processing) was fundamentally inadequate to explain subjective experience. He proposed an alternative framework drawing on quantum mechanics and field theory.

The philosophical arguments in the book were contested. But the personal arc was coherent: a man who had spent fifty years building silicon devices that executed instructions had concluded that what mattered most (awareness, experience, the sense that something is happening) was not reducible to any computation he could design.

He had also, by that point, received most of the credit he had long deserved. The Kyoto Prize in 1997, the National Medal in 2010, and the ongoing recognition of the Z80’s historical significance meant that his name was associated, in the historical literature, with the inventions he had made.

Dead End: The Calculator Market

The 4004 was designed for a calculator application that largely vanished before the chip reached commercial maturity.

Warning

Busicom, the Japanese calculator company that had contracted Intel for the chips that became the 4004, filed for bankruptcy in 1974, three years after the chip was announced. The consumer calculator market collapsed under the pressure of competition from Texas Instruments, Hewlett-Packard, and Casio, whose falling prices squeezed the smaller makers. Busicom’s business, selling programmable desktop calculators, did not survive it. Intel had negotiated back the rights to market the 4004 to other customers in exchange for a price reduction to Busicom (accounts put the money involved at $60,000); without that negotiation, the microprocessor might have been a proprietary calculator component rather than the foundation of a new industry. The original application that justified the 4004’s development was obsolete before the microprocessor revolution it enabled had fully begun.

The full story of how the microprocessor transformed computing is told in The Integrated Circuit Revolution. The Intel context is covered in Andy Grove and Intel and Gordon Moore and Moore’s Law.

📚 Sources