[Beowulf] how can I calculate peak performance of a cluster

Stuart Midgley sdm900 at gmail.com
Tue Sep 27 17:11:43 PDT 2005


I found everyones post interesting, but perhaps not answering your  
question.  The simple answer is to do the math.  A Xeon/Opteron/ 
Athlon64 can do two double precision floating point calculations per  
clock cycle.  Muliply up...  for 1 cpu

     2*3.4 = 6.8 GFlops
     2*3.2 = 6.4 GFlops
     2*3.16 = 6.32 GFlops

now multiply by the number of cpu's of each and you have your magic  
"peak" performance.

Processors like the Itanium2 and the IBM Power systems can do 4 flops/ 
cycle so get double the peak performance for the same GHz.

This is a very useful number.  We tell it to our funding body all the  
time indicating how much we manage to purchase with our measly  
Australian dollars.

There have been a lot of good papers published over the years on the  
topic of benchmarking.  For the true test of performance the best  
benchmark is your own code (or your users code) and how much effort  
you are willing to put in to optimise your code for a given system.

Most computational codes only see ~20% of peak performance.  The rest  
of the time the cpu is sitting stalled waiting on memory, network,  
disk etc.


On 27/09/2005, at 20:15, guren at be.itu.edu.tr wrote:

> Hi,
> I want to calculate peak performance of cluster which  HP HPC Cluster
> consist of different numbers of INTEL Xeon 3.4GHz EM64T , INTEL  
> Xeon DP
> 3.2GHz EM64T ,INTEL MP XEON 3.16GHz EM64T processors. I think it is  
> easy,
> but I couldn't find.many
> many thanks for help
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit  
> http://www.beowulf.org/mailman/listinfo/beowulf

Dr Stuart Midgley
sdm900 at gmail.com

More information about the Beowulf mailing list