Error: Can't find /dev/nvidiactl

I am just getting started with CUDA on Redhat 2.6.18.

A week ago, I compiled and tested all programs released with the CUDA_SDK.
They worked fine. The program deviceQuery gave correct information
about the GTX 285 that is in the machine.

Yesterday, while running my own (attempted) CUDA program, my monitor
and keyboard locked up and refused to respond. I did a hard reboot.

Now no programs from the CUDA_SDK will run!
The program, deviceQuery gives the error:

      NVIDIA: could not open the device file /dev/nvidiactl (No such file or directory).

My hardware is an MSI motherboard with GeForce 8200 chipset. This is for the monitor.
The graphics accelerator is a GeForce GTX 285. This is intended for CUDA.

According to the config file, SELinux is “disabled”.

How can I restore my machine so it correctly runs the CUDA_SDK tests
on the GTX 285?

Also, how can I configure things so the GeForce 8200 on my motherboard drives my
monitor, while the GTX 285 runs my CUDA programs?


I get that message when the X-server hasn’t been started.

On my gentoo box,
/etc/init.x/xdm start
will do the trick, as this will load the appropriate drivers and such.

I think one could shortcut the above with something like

modprobe nvidia
mknod xxxxx

but I’m not familiar with mknod’s syntax, so I can’t give you any more details on that.

Thanks Letharion. I’ll try that.

Also, google found a relevant page which follows:


Neat, there are the relevant mknod commands.

Then I learned something new today. :)