dual Xeon issue

Craig Tierney ctierney at hpti.com
Thu Jun 5 15:35:16 EDT 2003


On Thu, 2003-06-05 at 08:20, sfrolov at accufo.vwh.net wrote:
> Sorry for a noob-like question. We are in the process of buying a Xeon-based cluster for scientific computing (running MPI parallelized numerical models). We've benchmarked clasters offered by Microway and Aspen Systems and discovered that we are getting much better results using only one CPU per box (running the model on 6 boxes using one CPU in each box was about 50% faster than running it on 3 boxes using both CPUs). This was surprising to us since we've used to be limited by interprocessor communications, which should be a lot faster between CPUs in one box. Can anybody explain the reason for this and, more importantly, is there any way to improve this situation.
> 

You are probably pushing the memory bandwidth on
the motherboard.  On most Xeon motherboards, the memory
bandwidth can be exhausted by 1 cpu.  Rarely do
you get a 2x speedup when using the 2nd cpu.  Your results
are typical.

You could go with single CPU motherboards.  The new 800Mhz
FSB looks quite attractive.  However, the odds are that when
you include the extra interconnect and the other infrastructure
that the dual Xeon system will provide better price/performance.

It does all depend on your code.

I doubt that a Microway or an Aspen Systems cluster is going
to perform that differently if configured similarly.  Make
sure that Hyperthreading is off and you are using a good
compiler (Intel).  Look at what support you get for your money.  Things
are going to break, and knowing how the vendor is going fix your
problems is important.

Craig

> Thanks,
> 
> Sergei.
> 
> P.S. Can anybody recommend the best performance/price ratio cluster?
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
-- 
Craig Tierney <ctierney at hpti.com>

_______________________________________________
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf



More information about the Beowulf mailing list