I have a Dell 1950 server running 64-bit RH 5.5 with a 2.6.18 kernel. It is connected to a Tesla S870 GPU via PCIe. Today, I downloaded Cuda 5.0 and installed it on the server. I also installed the latest video driver. Upon reboot, I was able to build and run several of the sample applications. However, X windows would not start up because the driver was the Nvidia driver, not the expected Radeon driver. If I switch back to the Radeon driver, X starts upon fine but I can no longer communicate with the GPU. Is there some way I can load both drivers, using the Radeon for graphics and the Nvidia for the GPU, such that they can peacefully co-exist?