INTEL XEON vs AMD ATHLON
lindahl at keyresearch.com
Wed Jun 12 06:50:06 PDT 2002
On Wed, Jun 12, 2002 at 08:37:46AM -0500, Rocky McGaugh wrote:
> The second is its greater PCI bandwidth. The E7500 performs much better
> than the Athlon motherboards.
Please do not use my site for benchmark abuse. As it says on the
website, actual end-to-end Myrinet performance is limited by the
weakest link in the entire transmission/receipt process. Once PCI gets
good enough, it is no longer a bottleneck. If you're just looking at
the raw numbers, you're getting the wrong answer, and you're doing a
disservice to the community.
> The third is the availability of specialized compilers. This may or may
> not be a big win for you.
I've found that the Intel compiler generates code which runs pretty
well on the Athlon. The Intel compiler is also good at generating code
which is specialized to the P4's SSE2 registers, but that's sometimes
slower than ordinary code which does not use the SSE2 registers. I
find the compiler "issue" to be more FUD than real. What's your
experience with it?
More information about the Beowulf