HPC, OS X, Gotos considered non-harmful

Many thanks to Gaurav Khanna for compiling his HPC on OS X site. I'm currently in the middle of performance testing the latest crop of commodity 64-bit processors, and this site was of great help in the OS X LINPACK tests. I've also been corresponding with Kazushige Goto, as he is optimizing his BLAS libraries for the EM64T processor. I'm getting good results with his current pre-release library, and each revisi

Many thanks to Gaurav Khanna for compiling his HPC on OS X site. I'm currently in the middle of performance testing the latest crop of commodity 64-bit processors, and this site was of great help in the OS X LINPACK tests.

I've also been corresponding with Kazushige Goto, as he is optimizing his BLAS libraries for the EM64T processor. I'm getting good results with his current pre-release library, and each revision seems to eke out more ops from the new CPUs.

In a response to my recent article describing my 32-bit benchamarking of Intel's first foray into the x86_64 market, David T. Johnson asked:

So??? Which is faster, Nocona or Opteron? Your article keeps hinting at some sort of Opteron vs. Nocona comparison... Unfortunately, you never make any comparison between Nocona and Opteron. How do they compare on 32-bit performance??? You are the reviewer with the side-by-side test machines. Cannot you find a way to make a comparison?

Well, I'm deep into that at the moment. The issue at the time of publication was hardware availability. Comparing the new Dell 2800 with dual 3.6Ghz EM64T CPUs to the Newisys 4300 4P 2.2Ghz Opteron 848 lab server really wouldn't be a fair fight, even if I cut two CPUs out of the 4300.

At the moment, however, I have almost all the pieces in place to make that comparison, and I should think that it will be worth the wait.

Recommended
Join the discussion
Be the first to comment on this article. Our Commenting Policies