Re: Using 3 GPUs (but not 2) causes energy and temperatures to rise extremely fast in NAMD 2.12

From: Bryan Roessler (roessler_at_uab.edu)
Date: Thu Feb 16 2017 - 12:10:23 CST

It turns out I eventually experienced the same issue with two GPUs, it just
took much longer to manifest.

It looks to be a problem with using the useFlexibleCell or useConstantArea
parameters. I was using these as part of a lipid membrane simulation, but
have been experimenting without them and the results seem to be OK without
the associated temperature, energy, or artifact problems I was having with
them enabled in conjunction with CUDA2.

*Bryan Roessler | Graduate Research Assistant*
UAB | The University of Alabama at Birmingham
*uab.edu/cmdb <http://uab.edu/cmdb>*
Knowledge that will change your world

On Tue, Feb 7, 2017 at 7:55 PM, Bryan Roessler <roessler_at_uab.edu> wrote:

> I've been trying to track down why my simulations in NAMD 2.12 were giving
> me incorrect results and out of control temperature and energy outputs.
> Temperatures would rise to 330K+ in a few thousand steps and a large
> depression was forming in my solvation box.
>
> I've discovered that using 3 GPUs simultaneously in NAMD 2.12 causes the
> issue. Manually selecting one or two devices (it doesn't matter which
> device(s)) using +devices allows the simulations to proceed normally. I
> never had an issue with using 3 GPUs prior to NAMD 2.12 so it's not a
> hardware or power supply issue.
>
> Hopefully this will help you track down the bug.
>
> Thanks,
>
> Bryan
>
>
> *Bryan Roessler | Graduate Research Assistant*
> UAB | The University of Alabama at Birmingham
> *uab.edu/cmdb <http://uab.edu/cmdb>*
> Knowledge that will change your world
>
>

This archive was generated by hypermail 2.1.6 : Sun Dec 31 2017 - 23:21:04 CST