unable to fing cuda device as user


We are using NPACI rocks cluster 5.3 for a 3 node cluster and we have 8400 GeForce GS GPU’s on two compute nodes and one frontend without GPU. We installed the NVIDIA 256.35 drivers and we also installed cuda 3.0 on all the 3 nodes and we installed GPU SDK as user on all 3 nodes.

When we tried to run the sample device project as user we are getting the API error saying that there is no device that support cuda. When we install the GPU SDK as root and when we run the same device query it is showing the GPU and everything is working fine.

We edited the _bash_profile and .bashrc for both user and root.

Can anyone please help us how can we run device query as user and get the program running on gpu by cuda.

Please help me.

set permissions on /dev/nvidia*

Thank you very much for your reply. But when i installed the dev drivers and go to

cd /dev

ls -l | more

if u scroll down by hitting space bar , I am not able to see any nvidia* file or directory there.

I presume you are running without X11? If so, you have to have something create those device files at boot. It is explained in the toolkit release notes for Linux.