-
Conclusion/Epilogue
- Johns Hopkins University Press
- Chapter
- Additional Information
282 Conclusion/Epilogue Although the semiconductor industry had clearly established MOS technology by 1975, one could assess its impact in two different ways. If one were to focus on its present accomplishments, one would see a niche technology that had made some new things possible but that had had little effect on either the semiconductor or the computer industries. On the other hand, if one were to strain to look out into the future, one might see a revolution that was yet only dimly manifest. By 1975 the MOS transistor had not led to any fundamental reorientation in the semiconductor or computer industries. Although it had sent magnetic core memory down the road to obsolescence, this transition had been relatively easy for computer companies to make. IBM had incorporated MOS memories without fundamentally changing the way it developed its computer systems. Although the decreasing costs and increasing densities of MOS memories were responsible for the premature demise of Digital’s PDP-11 computer, Digital had been easily able to comprehend and rectify the problem. If one were to assess who had lost with the coming of MOS technology, ironically the biggest losers would be the MOS firms themselves. First there was General Microelectronics, which had made an early commitment to MOS technology but had not been 283 Conclusion/Epilogue successful. Then came Viatron and Cogar, followed by the firms that had been run over in the stampede to produce MOS calculators. Even by 1975, it was not clear whether the cumulative profits made in MOS technology offset the losses. Intel, which had taken MOS technology further than any other semiconductor company, had risen to fifth place in the semiconductor industry by 1975, an impressive achievement. But each of the top four semiconductor companies, Texas Instruments, Motorola, Fairchild Semiconductor, and National, still relied on bipolar technology for over two-thirds of their sales.1 But MOS technology had opened new markets for semiconductors, including calculators, memories, watches, and an assortment of low-cost, lowperformance applications. Through the proliferation of the calculator, the general public received firsthand experience with digital electronics and access to unprecedented computing power. Millions of Americans now wore digital watches. The combination of the MOS microprocessor and MOS memory would make it possible for thousands of Americans to have a computer in their homes. By 1979 the consulting firm Integrated Circuit Engineering made the heady assertion that the potential existed to put at least three microprocessors in every house, and an equal number in every car.2 A paper written in 1974 emphasized another aspect of MOS’s potential, spelling out unstated but common knowledge in the industry: the MOS transistor could be scaled, and these smaller and smaller transistors would lead to more and more powerful integrated circuits. Ironically, the paper came from IBM, which would be most affected by its implications. Robert Dennard and a group from IBM Research gave a detailed exposition of the principles of MOS transistor scaling, showing that as one scaled an MOS transistor by a constant factor, the delay of the transistor scaled down by that factor, while the power per circuit scaled downward with the square. (This meant that smaller transistors would be faster but consume roughly the same power per area.) Dennard’s group accompanied the analysis with experimental work making MOS transistors with dimensions as small as one micron. (At the time, IBM’s MOS memory chips in production had dimensions as small as five microns.) IBM Research was able to make such small devices through the use of ion implantation, a technique whereby dopant atoms were accelerated into the silicon rather than thermally diffused, allowing for very precise control. Although one could not yet make devices that small in a manufacturing environment (and much more [3.235.42.157] Project MUSE (2024-03-19 12:34 GMT) 284 T O T H E D I G I TA L A G E work was required to make that possible), Dennard’s group showed that it could be done. And given the economic incentives to produce larger memories and more capable integrated circuits, MOS technology clearly had a long run ahead of it.3 Science and Technology, Research and Development What does the MOS transistor say about industrial research in post–World War II America? Throughout the 1950s and 1960s managers’ thinking about research and development was shaped by Vannevar Bush’s linear model that new science would produce new technology. But the MOS transistor hardly followed...