Re: K6 scored higher on DOS applications when the PPro first came out
By the time K6 came out no one cared about DOS. K6 did very well on integer apps, particularly clock for clock, and not very well on floating point apps. PPro did very poorly on apps that included 16 bit code, but the result was mostly sales lost to Pentium, not to K6.
P4 is outperformed by Athlon on both integer and floating point, using "internet content creation" as a proxy for floating point and "business productivity" as a proxy for integer apps.
bapco.com
K6 was doing very well for AMD until they encountered first a yield crash, then later the chip stopped scaling (sort of like P3 has stopped scaling for Intel).
This is mostly a game of minor stepping changes since new cores come out very rarely. The most recent stepping for Athlon seems to have taken it from a range of 900/1100 to a range of 1100/1300 while the most recent stepping of P4 took it from a range of 1400/1500 to a range of 1300/1500 (but availability indicates yields are much better). Considering that P4 clock for clock is slower and requires an expensive dual channel memory subsystem to get what limited performance it does achieve, P4 is looking rather tired - and it's only 3 months old.
Something delayed AMD's 266FSB parts. It may have been the chips, it may have been the motherboards, and it may simply have been too much 200FSB inventory waiting to be cleared out. AMD can't afford to write off a quarter's production the way Intel has been.
Whatever the cause, the 266FSB parts and 1.3 clock speed is finally here. Considering that a 1 to 1.1 GHZ Athlon is performance competitive with a 1.5GHZ P4 when both are running the same OS ( bapco.com ), it wasn't a good time for Intel to have a stepping not pan out. Availability seems to indicate that PIII now has a sweet spot close to 933, and decent binsplits at 1GHZ, but 933 is now too slow for even a midrange system.
I would guess that Intel has already shifted to modified steppings that will show up in a month or two, which may go far towards solving their problems. But someday AMD may actually produce Palamino, and if Intel isn't way ahead of where they are now, it's going to get ugly for them. OTOH, Intel will get decent steppings of PIII and P4 out eventually, so AMD had better get Palamino figured out one of these days.
But in this latest lap in the Marathon of steppings, Intel has lost some ground. P4 should have blown Athlon back into the midrange until Palamino came out - but it hasn't. Now all Intel can do is pray Palamino is a dud, they have nothing to fight back with until their .13 is in volume, between 6 months and a year from now.
Currently, Intel is a sitting duck. The question is, can AMD get off a shot before the duck paddles away?
Dan |
