another question

[Keith Bierman fpgroup]

In my youth, I worked in a cap factory. It resembles the chip market
to some extent.

	b) For workstation/server chips, the life of  fab isn't very long, i.e.,
	people are always looking for the newest, zingiest process, so maybe
	you get to use that process for high-end chips for 18-24 months.

Yeah, but as mentioned later in your article, many times a fab lives
on doing more pedistrian work later on. Either derived from the
original fancy chip, or other stuff ... few companies really flush
*all* the investment in 24 months (and live to tell about it, round
after round ;>).

Not wanting to tweak the process is a good point. On some lines "we"
managed to get something like 99% yields ... but those were nice old
processes, and extensive binning was employed (virtually everything
got sold to *somebody* but the mil-spec stuff paid for all the costs,
the rest of the yield was gravy). Then there was everything else in
the shop; the more a line was jerked around, the lower the yields got.
It was quite predictable. *sigh*

	Under such circumstances, it is clear that quoted prices and actual
	costs may well be unrelated, and it is very hard to tell what's going
	on inside a captive fab.

Also it may be unreasonable to really burden the chips that way. The
alternative may be to get out of the biz entirely ;> But accounting is
an odd science at best, FASB's come and FASB's go ...

	uses MIPS R4000s for embedded avionics, for example].  PPCs are more
	reasonable for embedded control, although a lot of embedded could
	care less about floating point.

But a lot of embedded systems can benefit from it. I recall one missile
project where the old estimation/navigation system was fixed point.
Use of floating point allowed the software to be simpler, eased use of
a better algorithm and an order of magnitude improvement was
obtainable (I don't know what they chose to ship; I had no need to
know ;>). I've long thought the ES chip end of the market was ripe for
a floating point takeover. The 4 and 8-bit machines won't stay ahead
forever (at least one can hope).

Also DSP sells well into some embedded markets, not the fp we usually
think about, but not necessarily unrelated.

	I'm not sure I understand the long-term economics of all this; fabs do
	cost real money.

Not having a VAX follow on costs even more real money. What choice do
they really have? They *could* have focused on another round of two of
VAXen and begun the migration to merchant parts; but that would have
been a very bold (but depressing) move. 

>4) A few numbers:
>According to RISC Management (but cross-checked with otehrs, and probably
>close), in 1992:

>Intel i960	2095K units	(HP LaserJet; more in 1993)
>AMD 29K		 849K		(laser printers, more in 1993)
>SPARC		 320K		(more in 1993, mostly from Sun growth, I think,
				as Sun was something like 90% of SPARC w/s)
>MIPS		 291K		(about 90K systems, 200K embedded)
				Expect >1M in 1993, from OKI printers
>(these are merchant chips above).
>HP PA looked like in 70-90K range, and IBM in 60K range, in 1992,
>although I didn't have one singel source for those numbers.)


It would be interesting to see the predicted growth curves. I'd guess
that the i960 isn't going to see any huge upswing, nor the 29K, nor HP nor
SPARC for that matter. Given the large percentage of ES credited to
MIPS, it would seem like it is like to have a steep curve. If the
claims of PPC winning Ford are correct, and Apple's software story
works out, PPC ought to have a real killer slope too.