Advanced Micro Devices Inc (AMD)

[pgerassi]

Dear Chipguy:

But getting 2 4P 850 systems for EDA work versus the 1 2P IPF system will make the IPF 1/4th the speed for the same cost. What if work like data mining does very well on MPP systems. And in those siutations throughput means more than latency. Getting 6 answers in 24 hours is better than 1 answer in 16 hours for the same money.

BTW SPECfp does not really correspond to EDA performance as the EDA datasets are much larger. IPF wins SPECfp due its large caches (the 1.5MB L3 version is 19% slower per clock than the 6MB version). Once they run out, the higher BW (25.6GB/s vs 6.4GB/s) and lower latency (60ns vs 110ns) of 4P Opteron 850 really outdistances it.

Pete