Cirruslvr: IPC doesn't matter much, but only because the Athlon's IPC is already good compared to the competition. If the Athlon were merely equal to the P4 in all situations, then higher IPC would benefit AMD, but since it is already fine it doesn't matter anymore. So clock speed is what matters at this moment.
I agree with what you are saying... but only in a certain, relatively narrow regime. When the IPCs are roughly equal, clock speed is the all-important factor. However, it's fairly clear (or, at least, it's my claim) that if the IPC of one were… say… 10x the IPC of the other, the McMannis law of "MHz Sells" would break down. The question is if AMD can push the IPC high enough to accomplish this.
To some (admittedly very limited) extent, I believe they already have. I hear people around me talking about buying an Athlon over a P4 for performance reasons in much the same way I heard people talking about buying a PII over a K6.
I know you didn't argue against this, but high IPC alone hasn't made a desktop X86 processor successful, as Scumbria reminds us.
I believe Scumbria has at least seriously entertained (if one can do such a thing) the idea of introducing "P-rating" for the Athlon - at least I seem to recall him arguing the point. Certainly, in the extreme case of IPC disparity, something other than "pure" frequency is needed.
-fyo |
