Intel Architecture Group VP Bill Siu is a versatile man, evidently. At this week's Microprocessor Forum he seems to have on the one hand told journalists that the Pentium 4 won't be Intel's big earner until 2002, while on the other he says that the amount of emphasis Intel will be putting on the P4 in Q4 will mean there won't be a lot of demand for the 1.3GHz Pentium III.
It is of course painfully obvious how this could be, but it does point up how vitally important marchitecture is to Chipzilla right now. Actually being able to build the chips now would be the ideal solution, but in default of that it's a case of break out the smoke and mirrors again.
Intel is now able to achieve volumes on the 1GHz PIII, but following the embarrassing rollout/rollin again of the 1.13GHz version in July, Intel isn't going to try again with it until Q2 2001, when it's likely to be able to ship the chip in volume. That leaves a fairly large window during which nothing much exciting will happen at PIII level, so over we go to P4 marchitecture.
Intel is now rolling out the P4 next month, but it's a big beast, difficult and expensive to build, and initial yields don't seem to be good. Despite that Intel last week jacked up the initial clock speeds from 1.3 and 1.4 GHz to 1.4 and 1.5 GHz, and cut the price while it was about it. The rationale may of course be that if you can't make the volumes you might as well not make them at a higher speed, and make a lot of noise about them anyway.
The last months of this year are however likely to have a familiar feel to them, as Intel engages in propaganda wars with AMD on the basis of its latest short supply flagchip.
There's something else confusing about what Siu's been saying. He doesn't expect the P4 to be either Intel's major volume chip or its major revenue earner in 2001. Nor does he expect it to become the big volume one in 2002 - that however is when it'll earn Intel more than the PIII. That actually puts the P4 as the volume chip from 2003. Nevertheless, Siu has also been predicting that the P4 will move into the mainstream very quickly.
If you think about it very hard it's just about possible to figure out how something could be in the mainstream without being the major volume or major earning chip, but it's tough. Most likely Intel expects to get the P4 die size down, volumes up, and then try to sustain it as a high-margin earner for as long as possible before dropping it down to commodity levels.
But in addition to getting on top of its manufacturing problems, that also requires that Intel gets back to producing flagchips that clearly outpace AMD. Which is a toughie - in the interim, back to marchitecture and the shortages. ®