Albert, Jonathan Joseph is our friend? Who'd have thought? One part you didn't highlight:
We believe the significant challenge for Intel in the next 6-9 months is its
transitions from the old PIII architecture to the new P4 architecture. As Intel
management commented at the recent Developer Forum, the PIII was never designed
to run much above 1GHz. Largely for that reason, yields at that speed grade and
higher remain below average, and the 1.13GHz product was recalled. Meanwhile,
we do not expect the P4 to hit large volumes until mid-2001. It has two
problems: 1) the P4 is wholly dependent on RDRAM for its primary memory. Intel
bet heavily that RDRAM would be widely deployed by now, expecting it to be 20-
30% of the market in Q4. Instead, RDRAM, which is costly and does not show much
of a performance advantage, is only about one percent of the market currently.
Intel will not have its own SDRAM PC-133 chipset until mid next year, though it
may license Via to produce a chipset earlier. 2) the die size is large, and
some OEMs actually believe the P4 underperforms the PIII at certain speed
grades and in certain applications. For this reason, the P4 will probably not
be a compelling technology until the 0.13-micron process shrink ramps in volume
later next year. We believe the P4 will be a significant new architecture for
Intel, only that the transition will take 6-9 months to smooth out.
Once again, gotta get to Willy fades out into gotta get to .13um. But at least Intel is covering their bases there with plans to take p3 along for the ride. .18-.13 works out to a 50% area shrink, which still leaves Willy bigger than the current cumine die. Maybe THE WATSONYOUTH or some of the other process guys can take a guess at frequency scaling at .13um for p3 vs. p4?
Cheers, Dan. |
