TESLA Card with non NVIDIA graphic card

I recently purchased a TESLA C870 GPU card and I tried to install it on a Linux Machine. The drivers seems to be correctly installed, but the TESLA card is not working. It seems that there is a conflict on the X Server configuration.
Indeed, on my work station we have a PCI video card that is no NVIDIA to display on the monitor. Then when the X Server starts it launches the drivers for this card (which is a Matrox) and not the NVIDIA ones which appears to be necessary to run the TESLA card.
Can somebody help me solving this problem ? If possible, I’d like to avoid the answers like “replace your Matrox card by a NVIDIA one”.
Indeed, we only have one PCI Express x16 port, and when installing the TESLA card, the AGP port is no longer available.
I thank you in advance for your help.

Cheers,

Mathieu

If the nvidia binary driver is loaded in the kernel, then that’s all that’s really needed. As far as I can tell, the X driver only creates the /dev entries. So if you only have a single nvidia card, doing (as root):

/sbin/modprobe nvidia

/bin/mknod /dev/nvidiactl c 195 255

/bin/mknod /dev/nvidia0 c 195 0

/bin/chmod 666 /dev/nvidiactl /dev/nvidia0

should be enough to get CUDA up and running on the Tesla.

If you have more than one card, then create more device entry nvidia at character device major 195, minor , and set permissions on those.

Hope this helps.

There is a script in the release notes that will do this for you,
When you install the driver, do not configure X and then use the script to load the driver and make the proper entries.

I am also facing same above written problem . I have c870 telsa GPU card and I have matrox card for display . I have already follow given solution , but display (program like nbody simulation, particle etc.)is not coming . please help me out … If you have already solved this problem , please write me solution step by step .

For nbody simulation error coming like => freeglut : failed to open display .

I urgently need solution .

Thanks In Advance .

CUDA apps that use OpenGL interoperability must also run their display portion on an NVIDIA graphics card.