I view this year’s failure as next year’s opportunity to try it again. Failures are not something to be avoided.
You want to have them happen as quickly as you can so you can make progress rapidly. Gordon Moore
William Shockley co-developed the transistor at Bell Labs and then moved to California to establish the Shockley Semiconductor Laboratory. He recruited a number of promising graduates including Bob Noyce and Gordon Moore.
Shockley proved extremely difficult to work with; on one occasion he threatened to use a lie detector on the team over some relatively insignificant incident. Moore delicately described him as:
…a rather unusual personality, very bright, phenomenal physical intuition, but with relatively little idea how to work with people.’
When Robert Norton Noyce and Gordon Moore, plus six colleagues, left Shockley he called them the ‘Traitorous Eight’. They founded Fairchild Semiconductor and soon commenced a volume production of silicon transistors. Jean Hoerni, a Fairchild co-founder, developed the planar process to etch away the silicon dioxide coating using a photolithographic process. A film negative defined areas not to be etched, leaving the coating as an insulator. Where there was no image the coating was removed to become a conductor. Circuits were completed by interconnecting the conductors with aluminium wiring.
The first silicon mesa transistor, the 2N697, launched in 1958 and named for its raised flat structure. It was selected for the guidance system of the Minuteman missile. Noyce developed this process to create identical components on a single wafer.
He could etch diodes, transistors, resistors and capacitors. He filed for the patent for an integrated circuit (IC) in July 1959.
Jack St Clair Kilby at Texas Instruments used a different route, connecting capacitors, transistors and resistors with gold wiring that he called ‘flying wires’; this concept was announced in March 1959.
US courts agreed that Kilby and TI should have the patent for the IC but Noyce and Fairchild the patent for the manufacturing process.
1965 Moore wrote an article in ‘Electronics’ magazine on the speed of semiconductor development.
This became known as ‘Moore’s Law’. It suggested the equivalent number of transistors able to be included on a microchip would double every year. In 1975 he revised his estimate to every eighteen months.
Texas Instruments produced the first silicon transistor in 1954 (base zero for Moore’s Law); it was a single transistor. By the 1960s TTL quads/gates contained 16 transistors. In the 1970s, 8-bit MPUs had the equivalent of 4,500 transistors. In the 1980s, 32-bit MPUs held over 250,000 transistors and in the 90s, 32-bit MPUs had over 3m. 64-bit MPUs in the 2000s had almost 600m!
The MPU development path was not just about size, it was also about decreasing costs. Early Fairchild transistors sold for c$150; sales volumes drove this down to a unit price of a few dollars. Today for the same few dollars you can buy a 64-megabit dynamic random access memory with the equivalent of around 70 million transistors.
2005 Moore rescinded his ‘law’, saying exponential growth could not continue as the industry was approaching a fundamental barrier. Individual circuits would reduce to the size of atoms and at this size their behaviour and properties would change.