run cuda program without using its display, how? Can't figure out how to install cuda GPU

Hi, i’ll be expecting some kernel functions to be longer than 5 secs, while I still want to have X display, but I can’t find a way to install the GPU without using it as the display device (with my intel onboard graphics as the main display). When trying to install the NVIDIA driver, it says “You do not appear to have an NVIDIA GPU …”, is there a way to force the kernel to detect the GPU?

Has anyone done it before? or any suggestion?

I’m using Centos 5, GeForce 8800 Ultra, and a motherboard with intel onboard GPU

Which driver version are you trying to install?
Dos the 8800 Ultra appear in lspci output?

i’m trying to install the CUDA 2.0beta (as it seems to support RHEL 5.x better). 8800 Ultra doesn’t appear in lspci output as I’m currently not using it as my display.

Note: CUDA program worked before when I was using the 8800 Ultra as the primary display, but as I’ll be expecting some long kernel executions, I need to switch the display to use my onboard intel GPU, and then install the CUDA driver and toolkit.

if the ultra is not in the output of lspci, then it means that linux does not see it at all. Is it inserted in a 16x slot, and do you have all power cables attatched? You should also see the card in you dmesg. Or maybe some setting in BIOS? It can be that the BIOS prevents other graphics cards when you enable the onboard one.

i got everything connected as i mentioned above, i got it working when it’s the main display, but now the card isn’t even in dmesg. The BIOS may have disabled it, i’m not sure, doesn’t looked like I check that… i hate dell computers, so inflexible

This is either a BIOS or kernel bug, or perhaps the BIOS is intentionally designed to only allow a single display adapter. Regardless, this isn’t a CUDA/NVIDIA issue.