RE: NAMD CUDA on dual Nvidia 295 GPUs

From: Sadhu, Shubho (NIH/NCI) [F] (sadhusj_at_mail.nih.gov)
Date: Thu Jul 23 2009 - 09:11:48 CDT

Hi David,
I'm using the charmrun version for multiple GPUs and processors. Here's the command I use:
$NAMDHOME/charmrun ++local +p2 $NAMDHOME/namd2 +idlepoll md.in > md.out &
You could also use the +devices flag to specify certain GPU board(s) (you can find the numbers of the boards with the deviceQuery program in the Cuda SDK). I've found that if NAMD is using the same video card as the GUI, NAMD runs much slower.
Use "grep binding <NAMD log file>" to see which GPUs NAMD is running on. These lines are printed during startup.
Hope this helps,

Shubho
________________________________________
From: owner-namd-l_at_ks.uiuc.edu [owner-namd-l_at_ks.uiuc.edu] On Behalf Of David Chalmers [david.chalmers_at_vcp.monash.edu.au]
Sent: Thursday, July 23, 2009 8:20 AM
To: namd-l_at_ks.uiuc.edu
Subject: namd-l: NAMD CUDA on dual Nvidia 295 GPUs

Dear All,

We are trying out the CUDA version of NAMD (2.7b1 CVS version from a few
days ago) on a machine with a quad core i7 cpu and two Geforce 295 GPUs. We
are using Suse 11.1 with the most recent NVIDIA drivers and cuda toolkit
installed.

We are seeing a speed-up in NAMD running on a single core, but as far as we
can tell, the speed up is only consistent with a single GPU being used.
When we use the 'nvidia-settings' gui, the temperature of only one of the
four GPUs is rising much.

This leads me to a few questions:

Can I use all four GPUs?

Should I be using one, four (or some other number) of cores?

Which version of NAMD should I be using, multicore or the charmrun version?

Is there some way that I can better understand how many GPUs NAMD is using?

Thanks for any advice,

David
________________________________________________________________________

David Chalmers
Faculty of Pharmacy, Monash University
381 Royal Pde, Parkville, Vic 3053. Australia
________________________________________________________________________

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:53:04 CST