kash, The latest i saw on FPU benchmarks for PIV was pretty decent spec fp scores
Actually, P4 scores were below what an equivalently clocked P3 would get. SPECint scores were much better than P3, beating an equivalently clocked P3.
Intel will now try to emphasize integer code.
AMD with a HUGE MHZ lead is demand limited to 3M athlons, even with AMD doing 1.1/1.2GHZ OEM deals at $180/chip
Huh? AMD sold 2.1 million in Q3 with a 100 MHz lead and bad infrastructure support. Q4 vs. Q3 is worth at least 20%, passage of time is worth 15%, an extra 100 MHz for Q4 is worth 20% and better infrastructure is worth 20%. The result is 4.2M athlons (plus 3M Durons) in Q4.
Going from Q4 to Q1 I see a 20% drop because of the quarterly pattern, parially cancelled by the continuing 15% gain due to Intel inertia, a 0.5 M drop because of P4, so now we're down to 3.35M Athlons. But the 400 MHz gap results in a 20% increase and the even better infrastructure results in another 10%. So we have (4.2 x 0.8 x 1.15 - 0.5) x 1.2 x 1.1 = 4.4M athlons. The seasonal pattern for Durons would probably be a 30% drop, but a low cost infrastructure for Duron will help maintain the status quo on Durons (3M). So I see a total of 7.4M K7's in Q1, a marginal increase over Q4. Obviously, any K7 laptops would be gravy, since there will be none in Q4.
Its not clear to me how many of those folks in Q1 may jump to PIV -- say 1M.
Above, I modelled only 1M P4 sales in Q1, if they manage to ship 2M, then that would drop Athlon sales by an additional 0.5M to 6.9M.
In COMPUSA yesterday, I saw both 700 MHz and 750 MHz Duron machines for sale, but for ridiculously high prices. (Duron 700 = $900 with just a CDROM and no monitor.) I calculated that anyone could buy the parts for this machine for less than $450.
Petz |
