Skip to Content.
Sympa Menu

forum - Re: [abinit-forum] question on computer for abinit

forum@abinit.org

Subject: The ABINIT Users Mailing List ( CLOSED )

List archive

Re: [abinit-forum] question on computer for abinit


Chronological Thread 
  • From: Anglade Pierre-Matthieu <anglade@gmail.com>
  • To: lbo@ustc.edu, forum@abinit.org
  • Subject: Re: [abinit-forum] question on computer for abinit
  • Date: Mon, 19 Dec 2005 13:53:48 +0100
  • Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=TMG42sScgBD/HpiNHbgUOYnTLFrIV5YFqvJUU/qgfC1qnP9aUjLFhueJqBD5kQpIDnkCv4aWiiCy3y3jP4Ofs/rdMKz7MmY9OqPJXploj36CR/KOnB1OXjpHpv31nYp/WyNjrWtMu5PgZsxtusWKTqvQzSaCpd9To9Cpo01uaf0=

Hi,
I haven't test all these configuration yet here are some basic considerations:
-ECC memory is slower. It would provide you a little more safety
regarding some critical data. It is not mandatory at all for
scientific calculations.
-Since abinit uses heavily double precision computations with few
branches, the pipelines problem of p4 might not affect you too much
here. (Otherwise you would have easily chosen the AMD)
-Yet abinit like most dft code is very bandwith sensitive. That's why
the P4- 830 might not suit your goals. Indeed they share the memory
bandwith and it might have a strong impact on your computations.
-With Xeon you'll get half the computation power (only one core right?)
-Then my best guess is for the athlon X2.
-Note that for small calculations abinit is cach sensitive. You might
prefer some athlon 64X2 with large caches (4000+ is the same frequency
with twice more cache)
regards,
--
Pierre-Matthieu Anglade



  • Re: [abinit-forum] question on computer for abinit, Anglade Pierre-Matthieu, 12/19/2005

Archive powered by MHonArc 2.6.16.

Top of Page