Re: [AMBER-Developers] beta of pmemd.cuda.MPI available in git tree

From: Ross Walker <>
Date: Mon, 23 Aug 2010 11:51:44 -0700

Hi Dave,

> I'm still getting the following error on my S1070 system:
> cudaGetDeviceCount failed unspecified driver error

Try making sure the machine is in Runlevel 5.

init 5

Since if it was rebooted and only went to Runlevel 3 it probably won't have
loaded the graphics driver. This is a 'bug' in my opinion that NVIDIA needs
to fix since it doesn't make sense running servers at runlevel 5.

> I don't understand how to run or test cudaGetDeviceCount manually. I
> used
> the default locations for the driver installation, but I can't figure
> out
> from the installation script where the default actually is. So, I'm
> kind of
> stuck in knowing how to begin to debug this. Any suggestions are
> welcome.

The thing to run is the deviceQuery command from the cudaSDK.

This you can install and the make. If the deviceQuery command does not run
then that tells you it is definitely an issue with the driver not being

Although I suspect the 'init 5' will fix it.

You might also be able to do 'insmod nvidia' if 'lsmod' does not show it

All the best

|\oss Walker

| Assistant Research Professor |
| San Diego Supercomputer Center |
| Adjunct Assistant Professor |
| Dept. of Chemistry and Biochemistry |
| University of California San Diego |
| NVIDIA Fellow |
| | |
| Tel: +1 858 822 0854 | EMail:- |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.

AMBER-Developers mailing list
Received on Mon Aug 23 2010 - 12:00:06 PDT
Custom Search