Dear Andreas:
There are some problems with his reasoning.
1) AMD has five lines of processors, three of which are based on the same core (Tbird, Duron, and K75 Athlon (Slot-A)). The other two are (K6-2+ and K6-2). K6-2 is being phased out (K6-2+ may or may not be depending on how you interpret AMD's statements). Thus this leaves only Athlon based cores. Thus to compare Intel production which is comprised of five processors (Coppermine (includes Celeron II and small cache Xeon), Mobile Coppermine (includes mobile Celeron II), Celeron, Katmai, and large cache Xeon). The 700MHz figure is for Coppermines only. Mobile Coppermine, Celeron, Katmai, and large cache Xeons peak below this figure (Celeron and Katmai peak well below it). Thus one should only use Athlons to compare apples to apples.
2) Given that >= 1G Athlon production is >1 million a quarter and 3.6 million are scheduled to be sold leads one to the conclusion that 30% are at or above 1G. The typical speed curve tends to a Gaussian (Normal) curve. This puts the peak of all athlon production between 900 and 950MHz. Now this is substantially above the claimed 800MHz. Since Jerry stated that almost all 1G and above Tbirds are from Dresden and Dresden has about 40% of Austin in capacity, almost all Dresden Tbirds are at or above 1G. Thus, two things can be inferred, one is that Austin peaks at around 850MHz, and that Durons and Copper Tbirds are heavily under binned.
3) The gap is not 100MHz but could realistically be above 200Mhz. And when you compare Coppermine to Tbird, the gap is at least 300MHz and possibly more. There is a good chance that Copper Tbirds may have 5% or more above 1.4G and thus, matching P4 bins and production rates in Q4 (5% of 1 million is about 50K units and could be higher due to the Mustang family).
This is not looking good for Intel.
Pete |
