Hello Folks

I managed to install CUDA on OpenSUSE and I wanted to share my solution for those who have similar problems.

The Problem: Originally my system had the G03 version of nvidia display drivers from the OpenSUSE nvidia graphics repository. However, CUDA required the G04 version, so the CUDA installation replaced the original G03 version. This is where the problems started. After re-booting the graphics driver would not load and CUDA still didn't work.

The Solution: On my system I have multiple kernels from different kernel updates. The CUDA installation was placing the drivers in the wrong folder for my kernel. So basically the solution is to copy over the drivers to the correct folder. I have some bash commands listed below:

# Go to the directory where modules are installed
%> cd /lib/modules
# Lets look at all the kernels that I have
%> ls
3.16.6-2-desktop  3.16.7-21-desktop  3.16.7-29-desktop  3.16.7-48-desktop  3.16.7-53-desktop
# Lets see where the new nvidia kernels are installed
%> find -name 'nvidia.ko'
# Lets copy things over to the correct directory
%> sudo cp -r ./3.16.6-2-desktop/updates ./3.16.7-53-desktop/updates
# Lets install the kernels
%> sudo depmod -a
Question for experts:

How can I ensure that future kernel updates properly build and copy over the modules? I don't want to manually build and move around *.ko files for every kernel update. Clearly there is something wrong in the RPM or with Yast or something else. Could I just fix the relevant RPM and re-install? I just want to have a proper fix to this problem so Yast can manage my future updates seamlessly. Note I am running OpenSUSE 13.2.