[gmx-users] gromacs+CHARMM36+GPUs induces protein-membrane system collapse?

[Justin Lemkul]

On 10/31/18 8:36 AM, Ramon Guixà wrote:
> Hi again,
>
> Just confirmed that when I run the same tpr file for 500 ns using only
> CPUs, no problems arise, please check it out here:
> https://drive.google.com/open?id=1aI2wr2GvNjCfj4c4J0pmgCzV0ak5FRE6
>
> <https://drive.google.com/open?id=1aI2wr2GvNjCfj4c4J0pmgCzV0ak5FRE6>
> I mean, is this a bug? Most of my computer resources are GPU-based, is
> there anything I can change in the set-up or compilation so I can simulate
> using gromacs, CHARMM36, and GPUs?

Do you have access to a GPU with ECC? I wonder if this is a software bug 
or something going wrong with the GPU, like a bit flip. I've experienced 
this very rarely with consumer-grade cards like GTX 980, certainly not 
with the regularity you're seeing, though. There's clearly something 
wrong in your system to trigger such a sudden change. I pulled all the 
energy terms from your .edr file but it's hard to determine cause/effect 
in any of them.

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Assistant Professor
Virginia Tech Department of Biochemistry

303 Engel Hall
340 West Campus Dr.
Blacksburg, VA 24061

jalemkul at vt.edu | (540) 231-3129
http://www.thelemkullab.com

==================================================