Re: performance loss on GPUs

From: Aron Broom (broomsday_at_gmail.com)
Date: Wed Nov 28 2012 - 01:31:12 CST

Interesting. I've certainly never seen that with explicit solvent, but I
may have seen similar things with implicit. I'll have a look tomorrow on
some log files and send back a response. I have, however, only ever run
implicit simulations on a single GPU.

Also, how big is your system and have you tried +mergegrids on the command
line? For smaller implicit systems it seems to provide a bonus (in the
case of a 2k atom system, about 15%). Additionally, if you are making your
nonbonded cutoff very high (i.e. 999 angstroms to simulate no cutoff), try
setting "staticAtomAssignment yes" in the config file (I think this stops
needless pairlist reconstruction, and in the case where you want no cutoff,
can provide ~10% boost in the 2k atom case mentioned above).

~Aron

On Wed, Nov 28, 2012 at 2:13 AM, Thomas Evangelidis <tevang3_at_gmail.com>wrote:

> Greetings,
>
> Is it normal to see a performance loss on GPUs after a few hours of
> simulation. I am running an implicit solvent simulation on 2 Tesla M2070
> and 2 CPUs. At the beginning I got 0.0126392 s/step, but after 9 hours I
> get 0.0248662/step. Is this normal?
>
> thanks,
> Thomas
>
>
> --
>
> ======================================================================
>
> Thomas Evangelidis
>
> PhD student
> University of Athens
> Faculty of Pharmacy
> Department of Pharmaceutical Chemistry
> Panepistimioupoli-Zografou
> 157 71 Athens
> GREECE
>
> email: tevang_at_pharm.uoa.gr
>
> tevang3_at_gmail.com
>
>
> website: https://sites.google.com/site/thomasevangelidishomepage/
>
>
>
>

-- 
Aron Broom M.Sc
PhD Student
Department of Chemistry
University of Waterloo

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:22:18 CST