[Beowulf] dual core Opteron performance - re suse 9.3
diep at xs4all.nl
Tue Jul 12 02:22:11 PDT 2005
Good morning Don,
If you 'did get better performance', that's possibly because
you have some kernel 2.6.x now, allowing NUMA, and a new
compiler version of gcc like 4.0.1 that has been bugfixed more than
the very buggy series 3.3.x and 3.4.x
Can you show us the differences between the compiler versions and kernel
versions you had and whether it's NUMA?
Also how is your memory banks configured, for 64 bits usage or 128 bits
single cpu usage, or are all banks filled up?
When using all cpu's simultaneously obviously 64 bits is faster.
I know for Diep the NUMA part at dual opteron matters about 50% in speed.
For the version in question it was around 170-180k nps versus NUMA 255k nps.
Windows XP 64 NUMA reached 270k nps, but that's because the compiler is
superior to GCC.
Did you try opensolaris already and their free compiler tools?
I'm interested in speedcompares of it from objective (academic) sources,
what Sun will claim themselves i can guess :)
At 06:04 PM 7/11/2005 -0500, Don Kinghorn wrote:
>I'll start this thread to kill two birds with one stone ... gee that's not a
>very nice old saying is it :-)
>To answer the question about using SuSE 9.3. -- Yes you can use it in
>and yes, you can use it with dual-core Opterons too.
>However, I had to jump through some insane hoops to get it running on a Tyan
>S2891 board with dual-dual-core 2.0GHz Opterons, 4GB mem ....
>The results were quite satisfying though. We benchmarked our computational
>chemistry software (PQS) against a 2 node dual Opteron setup with the same
>general configuration. [2x Tyan S2875 and 2.0GHz Opterons 2GB/per board,
>We got better performance for most 4 process parallel (pvm) jobs on the
>dual-core system than on the 2 dual-node setup. The only slower jobs were
>jobs that did heavy disk i/o but this was expected. ... I have SATA-II
>on order :-)
>The main problem with the SuSE 9.3 install on the dual-core setup was that
>install didn't create the correct initrd to boot the system. I've worked
>around it using the SuSE 9.3 Live CD, (which does boot and run fine), by
>mounting the installed system and chroot'ing into it and fixing the initrd.
>>From now on I'll clone it.
>I'll be doing our cluster systems with the new Tyan board and dual-core
>Opterons with a configuration running SuSE 9.3 for our next hardware
>iteration starting in Sept..
>Hope this is helpfull info.
>Dr. Donald B. Kinghorn Parallel Quantum Solutions LLC
>Beowulf mailing list, Beowulf at beowulf.org
>To change your subscription (digest mode or unsubscribe) visit
More information about the Beowulf