Hi Micheal, Please, note the architecture and compiler also published (clik on the right side) and than you can see the used OS and compiler. Like the first one in the linked page: SOFTWARE -------- Operating System: SuSE Linux Enterprise Server 8 for x86 Compiler: Intel C/C++ 7.0 build 20021212Z and Intel Fortran 7.0 build 20021212Z I guess this is not a 64-bit compiler... Best wishes, Ödön P.S. By the way, I personally use Athlon64/SuSE 9.0 x86_64 and Gaussian compiled for 64-bit, but the facts are still facts. On Mon, 2004-02-02 at 11:11, Michael Matz wrote:
Hi,
On Mon, 2 Feb 2004, Odon Farkas wrote:
I did not say that the performance in the 64-bit environment is not good, just that the floating point performance is currently somewhat better in the 32-bit environment.
Hmm, hmm. You didn't by chance have tried to use GCC? I don't care about performance of code compiled by PGI. Anyway, right now this whole thing about fp performance on amd64 just being equal to pentium is just hearsay, as long as there is nothing demonstratable. At least our tests and also the published SPEC results tell another thing. Namely that the performance in SPECfp2000 of x86 class machines is just laughable compared to AMD64.
See http://www.spec.org/cpu2000/results/cfp2000.html .
I.e. something is strange on your side.
Ciao, Michael.