I’m installing CUDA5 on a GPU cluster instance (w/ m2050 Tesla Server model), but I guess the installation download .run file think the driver needs to be recompiled against the kernel source. I found the installation log said -kernel-source-path can be used to rebuild the driver (as this is a server system). I tried to set it up so the source tree along with the build configurations and install it with -kernel-source-path and point it to the same kernel source tree for the system that I installed. It doesn’t recognize the option. How can I CUDA 5 toolkit installed on a server instance like this? Or are there any download built for such GPU server Ubuntu 11.10 or 12.04?