I have installed the Nvidia Cuda Tool kit 5.5 and Samples codes in Ubuntu 13.1. I can compile .cu files and magmablas library for example.
I have an ASUS S550CB system with both an Nvidia GEForce 740MB graphic card and an Intel graphic card.
I would like to use the card to compute and not just compile Cuda code. Something that can be done in Windows even though we are forced to use Visual Studio which I find awkward and not flexible within this environment.
When I tried to install the deb package via apt-get command line in Ubuntu 13.1, it modified my settings. I was getting black screen and I was not able to log back in. My X server settings was corrupted and I was not able to turn the X-server back on. I had to re-install Ubuntu 13.1!
Is it possible to install the Nvidia Driver (and not just the CudaToolKit) without corrupting the X-server settings and getting black screen so that I can compile and run Cuda code?
if you use the runfile installer method, you can install the nvidia driver separately from the CUDA toolkit. If you run the nvidia driver installer separately, it will prompt at the end to modify the xorg.conf file. If you answer “no”, it will not make any changes and should not affect your X setup.
remove NVIDIA card
Reinstall OS, including GUI. Only the Intel card should be enabled for graphics at this point.
Download the latest driver for your GPU from www.nvidia.com, separately from the cuda toolkit.
Download the desired CUDA toolkit.
Add the NVIDIA card back into the machine.
Run the nvidia driver installer, selecting “no” when prompted to modify xorg.conf
Run the CUDA toolkit installer (.run file), selecting “no” when prompted to install the driver.(i.e. only install toolkit and samples)