[Q-e-developers] system size

[Kucukbenli Emine]

Dear Milinda, Laalitha

This question actually belongs to pw_forum mailing list (subscribe here : http://pwscf.org/mailman/listinfo/pw_forum )

Let me answer from experience,
Some years ago I have run calculations on a system of >1000 atoms of C H O N atoms (cholesterol molecular crystal)
from as little as 128 to as many as  1024 processors on an IBM SP6 machine and also on a Linux cluster.

I never had any problems with memory while running an optimization calculation, I dont think it ever exceeded 1.5Gb per processor.

I have just checked one of these calculations for you, 176 processors, the code prints as the largest occupied per processor dynamical memory as 961 Mb.

Note that back then we didnt have band parallelization.
I was also running my calculations with a very high plane wave cutoff for increased accuracy (required for NMR calculations I wanted to run afterwards) meaning that you can probably get away with less if you only want to optimize atomic positions.

Of course you can always use a less accurate but cheaper code first and fine tune your findings with Quantum Espresso later, but I found it was not necessary in my case, starting from experimental positions were enough.

best,
emine kucukbenli, postdoc at theos, epfl, switzerland



________________________________________
From: q-e-developers-bounces at qe-forge.org [q-e-developers-bounces at qe-forge.org] on behalf of Laalitha Liyanage [laalitha.liyanage at unt.edu]
Sent: Thursday, April 25, 2013 10:14 PM
To: Milinda Samaraweera; General discussion list for Quantum ESPRESSO developers
Subject: Re: [Q-e-developers] system size

Dear Milinda,

792 atom system is too large for Quantum Espresso.
You can try to use SIESTA http://icmab.cat/leem/siesta/

Laalitha Liyanage
Department of Physics
Central Michigan University
Department of Physics
University of North Texas

On 04/25/2013 01:01 PM, Milinda Samaraweera wrote:
> Milinda Samaraweera
> Computational materials chemistry and Computer programming
> University of Connecticut

_______________________________________________
Q-e-developers mailing list
Q-e-developers at qe-forge.org
http://qe-forge.org/mailman/listinfo/q-e-developers
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/developers/attachments/20130425/b2f4cc44/attachment.html>