- Bell LaboratoriesAt Bell, I dealt with the possible uses of Charge
Coupled Devices as memories. Because of their serial nature, their
niche in the hierarchy would have been similar to that of hard disk
memories, but with much less latency. As part of this work, I
established that CCD memories would never have enough cost advantage
over DRAM to gain a place in practical systems.
- Harris SemiconductorHarris
had a position in the market for Bipolar PROMs, and when I took over
the memory product development activity, Dave Taylor and I created a
new family of PROMs. Harris introduced the first member of that family
(512 x 4) in depth of the 1974/75 recession, and within a few years the
product family represented a very large fraction of Harris's commercial
business. Subsequently, we introduced a 1K NMOS memory, but that was
not really successful. On the other hand, with Chuck Gregory, we
created a line of synchronous CMOS SRAMs that were quite successful. A
big factor was the fact that Harris knew how to manufacture CMOS.
- Fairchild SemiconductorFairchild's MOS operation had three memory programs
going on when I joined. One was CCD memories, and while Fairchild had
built 64K CCD memories that worked pretty well, the old problem about
the cost and benefits had not changed. If anything, it had gotten
worse, because HDD technology had a steeper learning curve than IC
technology. We killed that one.
In the other areas of MOS memory, Fairchild was spread too thinly, with
activities in DRAM, EPROM and SRAM. The 1981 recession effectively
eliminated the DRAM program, because prices dropped well below costs.
Before I left the Fairchild memory activity, we worked bravely to enter
the high speed SRAM market, but the manufacturing resource in South San
Jose was clearly marginal. Yields were always too low.
- National SemiconductorAt
National, which had exited the DRAM market several years earlier,
Charlie Sprock had a dream that in a post Sematech world, he could
re-enter if he found the right partnership. Because of my Nagasaki
factory experience, I became the point man for a Mitsubishi
partnership. While we had some very productive plans, at the end of the
day, National could not enter at a competitive scale without building
an entirely new plant. We assessed the costs and risks, and the answer
had to be no.
After that, I managed, on a temporary basis, the high speed SRAM
activity in Puyallup, WA. Puyallup was serving Cray from a 4-inch wafer
fab. We qualified production, and ramped from zero to shipping a few
million dollars per month. However, having but a single customer doomed
that product line. No semiconductor business survives on one customer.
That was a major learning experience.