[CPMD-list] Thinking about buying an AMD Opteron system...

[Axel Kohlmeyer]

On Mon, 19 May 2003 13:24:02 -0400  mkosmows wrote:

> Dear CPMD community:
>  
> I am thinking about buying an AMD dual Opteron machine to run CPMD on
> and was wondering if anyone has any experience or opinions to share.  Most

well, if you want to run in 64-bit mode, you probably have to 'port'
cpmd to that platform (which probably is not a big deal, since you 
can 'steal' most of the flags from the alpha-linux configuration).

> importantly, I am hoping to find out for sure if an Opteron based machine will 
> be able to allocate more than 1 Gb of memory to a single process.  If I do 
> make the purchase, I would use the 64 bit version of Mandrake Linux 9.0 and 
> the 64 bit version of the PGI compiler.

well, you also need a very well optimized LAPACK/BLAS implementation.
e.g. for the amd athlon platform this can mean a 2x speedup for CPMD runs
(up to 8x for DGEMM). AFAIK, there is no such optimized LAPACK/BLAS
available (yet). you should check the ATLAS homepage 
<http://math-atlas.sf.net> to be sure. 

>  
> Thank you for any advice you all might have,

since cpmd does distributed memory parallization, you should also
check, how the opteron machine compares to a small dual-athlon
(with not too fast FSB266 cpus) cluster. in combination with
dolphin/SCI interconnect we have found nearly ideal scaling
and extremely high reliability. the latter cannot be expected
from a new platform (once you got the code actually working).

on the other hand: _if_ you get it to work (and don't have to
worry about wasted money), it probably will be a very nice machine.
the technical specs indicate, that the opteron cpu should be much
better suited for cpmd than current ia32 cpus.

but then again: if money really is not an issue, better get a
power4 or an alpha-cpu based workstation. those are known to
work, well optimized and also 64-bit.

regards,
	axel kohlmeyer.

>  
> Mark Kosmowski
> 
> Chemistry Department
> Syracuse University
> mkosmows at syr.edu
> 
> _______________________________________________
> CPMD-list mailing list
> CPMD-list at cpmd.org
> http://www.cpmd.org/mailman/listinfo/cpmd-list
> 

--

=======================================================================
Axel Kohlmeyer      e-mail:  axel.kohlmeyer at theochem.ruhr-uni-bochum.de
Lehrstuhl fuer Theoretische Chemie          Phone: ++49 (0)234/32-26673
Ruhr-Universitaet Bochum - NC 03/53         Fax:   ++49 (0)234/32-14045
D-44780 Bochum                   http://www.theochem.ruhr-uni-bochum.de
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.