By: Unununium (none.delete@this.none.com), July 25, 2012 5:55 pm
Room: Moderated Discussions
This chart was a bad idea when RWT published it years ago, and it's still a bad idea today. It's simplistic to the point of being wrong. Peak flops/mm and Peak flops/W don't matter. Otherwise, we would have seen widespread Silverthorne adoption in HPC after RWT crowned it king last time - which never happened.
What matters is delivered performance on applications people care about. This link at least attempts to address that question. The more interesting questions of delivered performance/$ and /W are harder to get at, but are the things that really drive decisions about which architecture to use. Peak flops/mm and Peak flops/W is marchitecture - the GHz wars all over again.
HPC large system performance
What matters is delivered performance on applications people care about. This link at least attempts to address that question. The more interesting questions of delivered performance/$ and /W are harder to get at, but are the things that really drive decisions about which architecture to use. Peak flops/mm and Peak flops/W is marchitecture - the GHz wars all over again.
HPC large system performance
