no GPU found when use SSH from another computer

Accelerated Computing CUDA CUDA Programming and Performance
1

I use Fedora11 and do scientific computation on GPU(GTX480).
Since several month ago, I cannot find any GPUs on the other computer by using SSH,then I found two way to solve this problem.

  1. go to that GPU computer and log in (or then “switch user”), then go back to my own computer.
  2. if nobody locally use the GPU computer, I can remotely use it without any problem. If anybody log in locally (I mean sitting in front of the GPU computer and use it), the GPU will “disappear”.

by the way, I use an additional graphic card to connect the monitor, the GTX480 are only used to do computation.

Who know why?

thank you

2

Did you set the rights on the devices (/dev/nv*) correctly, i.e. rw to all users? Also compare Linuxe release notes.

3

Did you set the rights on the devices (/dev/nv*) correctly, i.e. rw to all users? Also compare Linuxe release notes.

4

This the output of “ll /dev/nv*”

crw-rw-rw- 1 root root 195, 0 2010-08-18 09:23 /dev/nvidia0

crw-rw-rw- 1 root root 195, 1 2010-08-18 09:23 /dev/nvidia1

crw-rw-rw- 1 root root 195, 2 2010-08-18 09:23 /dev/nvidia2

crw-rw-rw- 1 root root 195, 3 2010-08-18 09:23 /dev/nvidia3

crw-rw-rw- 1 root root 195, 4 2010-08-18 09:23 /dev/nvidia4

crw-rw-rw- 1 root root 195, 255 2010-08-18 09:23 /dev/nvidiactl

crw-r----- 1 root kmem 10, 144 2010-08-18 09:23 /dev/nvram

Is there something wrong?

5

This the output of “ll /dev/nv*”

crw-rw-rw- 1 root root 195, 0 2010-08-18 09:23 /dev/nvidia0

crw-rw-rw- 1 root root 195, 1 2010-08-18 09:23 /dev/nvidia1

crw-rw-rw- 1 root root 195, 2 2010-08-18 09:23 /dev/nvidia2

crw-rw-rw- 1 root root 195, 3 2010-08-18 09:23 /dev/nvidia3

crw-rw-rw- 1 root root 195, 4 2010-08-18 09:23 /dev/nvidia4

crw-rw-rw- 1 root root 195, 255 2010-08-18 09:23 /dev/nvidiactl

crw-r----- 1 root kmem 10, 144 2010-08-18 09:23 /dev/nvram

Is there something wrong?

6

Being fedora, I would guess it is something to do with SElinux. I work in the exact scenario you are describing every day in a Ubuntu box, and have never seen anything like that. Check the kernel ring buffer and syslog and see whether there are any errors being reported.

7

Being fedora, I would guess it is something to do with SElinux. I work in the exact scenario you are describing every day in a Ubuntu box, and have never seen anything like that. Check the kernel ring buffer and syslog and see whether there are any errors being reported.