Bill - Re: "Could they not introduce test pieces(smaller/cheaper, but worked on just the same to vary a parameter fro that piece only?) here and there in the run, and glean feedback from them"
You are probably referring to test structures which can be used on short-loop experiments to check out various process steps - I would certainly hope AMD is doing this. If not, they would be the only semiconductor manufacturer not to use this technique. AMD can't be that naive so I assume they are actively making use of these techniques.
Re: "..You mean that AMD is acting like Goebbels, with the "big lie",
concealing the potential fact that there are no buyers "
Ahh.. my old buddy Goebbels!
Concealing the big lie? AMD has used this technique more than once.
I refer you to AMD's May 1995 announcement that the already late K5 would be delayed for "testing" purposes. That fooled nobody. The part was too slow and then AMD admiited they would have to shrink the K5 from a 0.5 micron process to a 0.35 micron process to get the speed improved.
Five months later AMD removed all doubt about their diffculties and announced that they were going to buy NexGEN and trash their internal K6 (not to be confused with the NexGEN designed K6 shipping today) development.
More recently, re-read the September 3, 1997 press release from Sanders describing an upcoming "minor" loss in Q3 due to a yield problem on the K6. The yield problems were pretty much resolved and AMD was then (According to AMD) producing up to 150,000 K6/week. The minor loss turned out top be $67,000,000 (before a tax credit).
Now do the math - 150,000 x 13 (Weeks/quarter) = 1.95 million K6's/quarter.
Well, on November 6, AMD announced that they were not going to meet their new goal of 2 million K6's shipped in Q4 - and I don't think they were talking about missing that goal by just 50,000 K6's!
Paul |
