Robert Noyce And The Road To El Dorado
Why the inventor of the silicon chip deserves more credit for the microprocessor
“Noyce knew exactly what he possessed in this integrated circuit, or microchip, as the press would call it. Noyce knew that he had discovered the road to El Dorado.”
Tom Wolfe: The tinkerings of Robert Noyce
“Don't be encumbered by history, just go out and do something wonderful.”
Robert Noyce
The story of the Intel 4004 has been the subject of much debate. Was the 4004 the first microprocessor? What was innovative in its development? Who deserves credit?
The 4004 was a success despite major missteps. It succeeded because of the determination of a small group of engineers. But it also succeeded because of Robert Noyce.
Noyce and Gordon Moore founded Intel in 1968. Noyce invented the silicon monolithic integrated circuit, or “silicon chip”, whilst at Fairchild Semiconductor. Moore was already famous as the author of Moore’s Law, his prediction that the complexity of semiconductor components for a given cost would roughly double every year1.
Their strategy at Intel was to design and manufacture advanced integrated circuits in large volumes. Initially, this meant memory chips for the makers of large mainframe and smaller mini computers.
The story of the 4004 began when Noyce was visiting Japan in April 1969. He was approached by Busicom, a small but ambitious manufacturer of calculators. Busicom had designs for several complex chips (at least seven) that would power a new range of desktop calculators. Could Intel, with its leading technology, manufacture them?
Noyce assigned Marcian ‘Ted’ Hoff to liaise with Busicom. Hoff soon saw that producing the Busicom designs would be unattractive for Intel. He started to develop the outline of an alternative: a design based on a smaller number of chips that would be reprogrammed to perform other tasks. Now assisted by Stan Mazor, the two worked to confirm that the design could meet Busicom’s requirements. The design was simple enough that Hoff believed that the main control functions of the computer could be incorporated into a single chip. That chip would become known as a microprocessor.
Noyce encouraged Hoff to continue developing his ideas. Noyce estimated that each set of the original Busicom chip designs would cost over $300 to make. He wrote to Yoshio Kojima, the President of Busicom. Soon the Intel alternative was presented to Busicom.
After considerable debate Intel persuaded Busicom management, on a visit to California, to adopt the new design. The contract was signed in October 1970 and the Busicom team headed home. Intel would make 60,000 sets of the new designs for a price of $60 each. Disappointingly for Hoff though, Busicom would keep the rights to the new Intel design.
But Intel’s memory business was struggling so a new project would not be popular. Noyce visited operations director Andy Grove. Sitting on the corner of Grove’s desk, Noyce casually announced “We’re starting another project.” Grove wasn’t happy but the project continued.
Except it didn’t. No significant further work was done on the project for six months. Intel simply didn’t have anyone who could turn Hoff’s outline designs into working silicon.
So when Federico Faggin, a young Italian engineer, arrived at Intel in April 1971 he found little more than Hoff’s outline. Faggin was recruited from Fairchild with the vague promise that he would be working on a project with ‘lots of logic’. Initially, delighted at the prospect of working on such an ambitious design, he was less happy when he learned about the state of the project. ”After one day on the project, I was six months behind.”
To make matters worse, a Busicom engineer, Masatoshi Shima, was arriving from Japan the next day. Faggin met Shima at the airport and drove him to Intel’s offices before showing him what he had on the new designs.
This was not what Shima expected. "I came here to check. No good. This is just idea. You bad. You late." Over the next few days, Faggin calmed Shima down, explained the situation and then persuaded his visitor to stay and help deliver the project.
Faggin was an ideal choice for the project. He had developed, whilst at Fairchild, silicon gate technology, which would be needed to build the microprocessor. He also understood computer design having previously developed a computer for Olivetti in Italy.
Working with Shima and Mazor, Faggin worked through the development of the four chips. Two were memory chips. One handled communication with the calculator’s keyboard, printer and display. Finally, there was the microprocessor, by far the most complex of the four.
The development process included laying out the design of the chips on large sheets of ruby-red Rubylith. Noyce would wander into the layout room, cigarettes in hand, and join Shima in checking the layout.
Finally, in December 1970, the first wafer containing the last of the four chips, the microprocessor, arrived on Faggin’s desk.
He placed the wafer on his test equipment. First chip: no response. Second chip: no response. Finally, he realised that the manufacturing team had omitted one of the stages in the manufacturing process.
A second, corrected, set of samples arrived in January 1971. They worked the first time with just some small bugs to fix. By March the chips started production and shipping to Japan.
In a telephone call with Shima in May, Faggin discovered that Busicom was struggling financially. Sensing an opportunity, Faggin and Hoff persuaded Intel’s management to buy some rights to the design. Intel would be able to sell the microprocessor for use in devices other than calculators.
Faggin decided that the chips should be named to emphasise that the four chips were designed to work together. They became the 4001, 4002, 4003 and finally, the microprocessor was named the 4004.
Intel launched the new chips with a double-page advert in November 1971. The portentous headline ‘Announcing a new era in integrated electronics’ was, for once, justified.
The 4004 soon found its way into cash registers and pinball machines. It wasn’t a huge commercial success but it established Intel as a leader in an exciting new market. The team, led by Faggin, iterated with an enhanced 4004, the 4040, and then much more capable designs, the 8008 and then the 8080. The 8080 was extended and enhanced to become the design that powers the majority of personal computers and servers today.
The 4004 was a team effort. It wouldn’t exist without Hoff’s original idea, Shima’s skills or Faggin’s hard work, drive and expertise in silicon gate technology.
And of course, it wouldn’t have existed without Noyce. Not just his creation of Intel. It was his original contact with Busicom. The encouragement he gave to Hoff to pursue his ideas. The visits to the layout room. Talking to Andy Grove. Above all his vision for Intel and his determination to take risks to achieve that vision. One step on the road to El Dorado.
Notes And Further Reading
So many words have been written about the 4004 in the fifty years since it was launched that I hesitated to add any more. Although mine is not a unique perspective, Noyce’s contribution is often minimised or omitted.
For a slightly more extensive account. I’d recommend the IEEE Article:
https://spectrum.ieee.org/chip-hall-of-fame-intel-4004-microprocessor
The recording of the thirty-fifth anniversary of the 4004 event at the Computer History Museum is essential viewing. Faggin’s grasp of the detail thirty-five years on is very impressive as is the fact that he gives credit, by name, to a number of the 4004 team.
Finally, for a short portrait of Noyce from an outstanding writer, Tom Wolfe’s Tinkerings of Robert Noyce is a great read.
https://web.stanford.edu/class/e145/2007_fall/materials/noyce.html
Photo Credits
Robert Noyce and Gordon Moore:
By Intel Free Press:
https://www.flickr.com/photos/intelfreepress/8268686452/sizes/o/in/photostream/,
CC BY-SA 2.0:
https://commons.wikimedia.org/w/index.php?curid=27929328
Moore would later revise his prediction to a doubling every two years.
Do TCP and UDP defaulting to port 8080 have anything to do with Intel naming their first successful server architectures 8080? Or is that just a coincidence?