As time goes on, AMD is beginning to look like some impotent Ottoman sultan. He and his minions may pronounce his virility to the world, but his harem is saying something else.
A few weeks ago, Cray told its shareholders that its revenues would be hurt by the lack of expected K10 processors. Yesterday, Soitec, the company that makes strained silicon wafers for AMD, announced that it expected considerably lower revenue the next twelve months, which appears to be due to AMD.
Yesterday, Digitimes said that the Taiwanese mobo makers were cutting back on planned AM2+ mobo production because AMD had told them Phenom would be delayed.
AMD denied this, or rather said something that sounded like a denial:
“However, AMD has responded to this report by stating that the company’s official launch schedule (bolded words our emphasis) for Phenom remains the second half of 2007 and this schedule has not changed. The company added that it has not contacted motherboard makers concerning a delay to the schedule.
Note the different wording. The Taiwanese sources are really talking when serious production starts, while AMD is talking about when the official launch happens.
If AMD can make and release a relative handful of slow Barcelonas in August and “launch” in September, there’s little reason to doubt that they could make a relative handful of slowish Phenoms and “launch” that towards the end of the year, with the real deal not happening until next spring.
The real revelation in the Digitmes article, though, isn’t “when” but rather “what/how”:
“AMD intends to modify its process technology to increase the yield rate and frequencies of Phenom processors in order to improve the competitiveness of the CPUs against Intel’s Penryn family, said the sources.”
That phrase covers a lot of ground, and while I could be wrong, “modify process technology” sounds like a good deal more than “come up with a new stepping.” As a bonus, it also confirms AMD’s current problems: speed AND yield.
What it really sounds like is that AMD gambled, lost, and now has to redo it, like Thoroughbred A and the initial ClawHammers. Thoroughbred A needed extra metal layers; Hammers needed extra IBM help with SOI. K10 will probably need something like that, not a quick fix, but not a complete do-over, either.
It’s also probably very safe to say that like Thoroughbred A and Clawhammers, overclockers ought not be the first kids on their block to get the first generation of K10s. Better to wait until spring.
Yes, there’s no guarantee this time schedule can be kept, or what results it will yield, though AMD does have a good track record of getting it right the second time.
The real question will be: Can AMD hang on long enough for these revised chips to come to their fiscal rescue, or will they burn up all their cash and credibility first?