Elmer Re.... I ran the numbers and showed you how AMD had yield problems. <<<<<
You ran what numbers? If I recall, your number crunching involved assuming AMD was running at all out capacity. I showed you from the CC that Hector specifically says, Dresden wasn't running at capacity. Your assumptions were wrong, therefore your results were wrong. In post 75489 you said this. There was more demand than Intel could supply in Q4 and still AMD couldn't sell out much more than half Fab30s capacity.<<< Okay, now that you admit that AMD was producing at half capacity, why not crunch the numbers again, using half the wafer starts and see what your yield figures are. That is the point. As for Intel, in your zeal to make Intel look good, (How you can possibly figure that Intel would look good having problems meeting demand during a recession, is beyond me), you also stated that Intel was producing at maximum capacity. You know how to run the numbers. Run the same formula for the 2 fabs producing P4s using their maximum wafer start no.s, which is what you are claiming Intel is doing, and see who has the better yields/wafer.
. You need to be able to do that before you can claim Intel has a yield problem.<<<<<<
You will please note that I stated that either Intel has a yield problem or a big die problem, which is causing low yields/wafer.
But that isn't Intels only problem with P4. If Dan3 and THG are correct about low memory speeds sharply reducing scaling performance, then Intel has a bigger problem, for as the speed goes up, in order for P4 to get higher performance, high speed, exotic,expensive memory, as in Rambus 1066 or better, will be required; meaning P4 will go from mainstream, to being a niche chip, because of the extra costs. The higher IPC, Hammer will operate better on regular DDR memory, and unless the situation really demands it, most will want to forego the extra costs. How many secretaries really need exotic memory ? |
