Re: benchmarking of NAMD: sustained performance (TF)

From: Axel Kohlmeyer (akohlmey_at_cmm.chem.upenn.edu)
Date: Thu May 29 2008 - 09:05:40 CDT

On Thu, 29 May 2008, satya work wrote:

SW> Dear NAMD users
SW>
SW> We are in the process of buying a 5-7 TF cluster and NAMD is one of the
SW> codes that is routinely used in our group. Are there any benchmarks
SW> available for NAMD performance in terms of flops for various
SW> machines/vendors? I have looked in the mailing list and also on the net,
SW> but could not get the right information. Most of the benchmarking results I
SW> found are in terms of seconds/MD step. I am interested in sustained
SW> performance of NAMD in terms of Teraflops (on various architectures) and I
SW> am wondering if such information has been compiled.

dear satya,

can you please explain what that would accomplish
on top of what you already know?

you can, in principle, easily convert that yourself!
all you have to do is to find out how many floating point
operations (=FLOP) are needed for one md step for a given
system and then multiply it with the inverse of seconds/step.

since for application performance it does not matter
how many flop/s an MD steps actually constitutes, but
how many of them you can do on machine a) in comparison
to machine b) for a realistic system.

cheers,
   axel.

SW>
SW> -Thanks
SW> Satya
SW>

-- 
=======================================================================
Axel Kohlmeyer   akohlmey_at_cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:49:30 CST