[Beowulf] trouble with linpack benchmarking - regd.

Greg Lindahl lindahl at pathscale.com
Mon Nov 21 10:54:45 PST 2005

>            Iam trying to benchmark a dual processor Xeon (nocona) 
> 3.6Ghz machine. I am successful in runnning the code. but getting very 
> low results with n=1000 linpack code. Iam getting a 260 mflops where 
> prior results submitted for the same processor show a performance of 
> around 3000 mflops.

You are comparing the performance of Fortran source to an optimized
assembly-language library.

-- greg

More information about the Beowulf mailing list