From: Bennion, Brian (
Date: Tue Aug 19 2014 - 11:41:12 CDT

Thank you John,

The VMDCUDADEVICEMASK 0x1 allowed vmd to start.
I tried to remove the bad card once and the other card didn't work when it was alone. I am not very knowledgeable when it comes to the SLI configuration my machine came with.

I will have my boys look into it more.

From: John Stone []
Sent: Tuesday, August 19, 2014 9:34 AM
To: Bennion, Brian
Subject: Re: vmd-l: vmd hangs at creating CUDA pool

  If you get a CUDA hang, the "disabled" card may not be disabled "enough",
to prevent CUDA from seeing it and trying to use it. You can either set
the environment variable VMDCUDADEVICEMASK to either 0x1 or 0x2 to force
VMD to use only the non-broken device, or you can set VMDNOCUDA to 1 to
prevent VMD from using both devices. Let me know if that lets you get
past the problem. The best thing would be to completely remove the failing
GPU from your machine so it cannot possibly interfere with your software use.

  John Stone

On Tue, Aug 19, 2014 at 04:11:17PM +0000, Bennion, Brian wrote:
> Hello
> Recently I had a failure in one of my quadro 5000 cards. It has been disabled but it was previously in a SLI configuration with my second quadro card

NIH Center for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801           Phone: 217-244-3349