Re: AMD has quite a track record when it comes to competing against Intel.
Mary,
Intel has (or had) a business where it makes parts for $25 and sells them for $200.
A business like that is going to attract new entrants. Intel has fought off several attempts to enter this fabulously lucrative business over the past 20 years. This time, thanks to Intel's concurrent love affairs with Rambus and VLIW, it let its business slide a little too long and it has lost its monopoly. If P4 had been a real home run, Intel might have been able to return to the earlier status quo, but P4 is a mixed bag.
P4 has high nominal MHZ - terrific for marketing. But P4 is fast only on synthetic benchmarks and a very, very, limited number of applications. Supposedly the day will come when most software has been rewritten to make P4's performance better, but such a day is many years away. The tools (like compilers) have to written (which is going on now). Then those tools have to become the preferred tools for developers, then applications have to be written (or rewritten) using those tools, then those applications have to enter the marketplace, then those applications have to become most of the application in current use. It's going to take quite a while, and may never happen if other chips are available that run all applications well - both mainstream ones and specially compiled ones.
P4 is also much more expensive to manufacture than its competition (both internal and external) and requires expensive infrastructure support as well.
Had P4 been able to run current programs twice as fast as existing chips, or been no more expensive to deploy than existing chips, it might have done what Intel needed. But P4's performance is downright sad on almost all software ever written and it is horribly expensive compared to the competition (I don't mean expensive at retail, I mean expensive to build - which will put some very hard limits on its ultimate market).
If Intel's shrink to .13 goes quickly and smoothly, and its transition to copper goes quickly and smoothly, and its transition to low-K wafers goes quickly and smoothly, and they are able to fit in the pieces of the P4 they removed to fit it onto .18 die without bringing it back up over 200mm2 on .13, and the design proves to scale well on the 3 ways new process, then and only then does Intel have a chance at regaining its prior status.
But for that to happen, it will be necessary for AMD to have problems with its somewhat simpler transition to .13 SOI from a .18 process that already uses copper. AMD will also have to run into problems with its less complex transition to the hammer series of chips. Because even if all goes perfectly for Intel, it probably won't be able to dominate a 64 bit Hammer on SOI that is around 100mm2.
Regards,
Dan |
