Linux box with two GTX680's, /dev/ files built six minutes apart

Hi -

I’m running two GTX680s in an Ubuntu 12.04LTS box. After a reboot, I noticed I could use one card immediately but not the other. nvidia-smi gave the error “could not open device file /dev/nvidia1”

I took a look at /dev/nvidia0 and /dev/nvidia1 … their timestamps are six minutes apart.
crw-rw-rw- 1 root root 195, 0 Apr 18 14:03 /dev/nvidia0
crw-rw-rw- 1 root root 195, 1 Apr 18 14:09 /dev/nvidia1
crw-rw-rw- 1 root root 195, 255 Apr 18 14:03 /dev/nvidiactl

Syslog didn’t catch anything of interest. All of the nvidia entries match the timestamp of /dev/nvidia0. No explanation of the delay there.

| NVIDIA-SMI 4.310.32 Driver Version: 310.32 |

Should this delay worry me? Or is it just normal settling time?

Thanks in advance.

It might be that the first card has an X server running on it, and so it is getting initialized (by X). The second card may be getting initialized some time later. This is really a function of how those /dev/ entries are getting set up.

  1. X can do it.
  2. You can do it manually with for example:
    sudo nvidia-smi -a
  3. You can create a startup script to create the entries automatically:

http://www.resultsovercoffee.com/2011/01/cuda-in-runlevel-3.html

In any event, if your setup is otherwise functional, you should be able to rectify the nvidia-smi error by running it as root:

sudo nvidia-smi -a