Dell R820 + P1000 (times two) CUDA headless install

Hi,

I have a headless Dell R820 with two nVidia P1000’s (in the garage… its loud)

Most of the installation instructions for CUDA assume a XServer will be used… is there a method to install CUDA from ssh for a headless configuration? Or do I have to pursue a XServer install and disable it?

With all the GPU farms out there this has to be a common thing.

I am running Ubuntu 18.04.

Any help would be appreciated.

Guy

I wouldn’t know what the xserver has to do with cuda install.

  • install the driver from repo (sudo apt install nvidia-driver-430)
  • download the cuda .deb
  • add the repo to your system (first three steps from install instructions on download page)
  • don’t install cuda
  • instead, run sudo apt install cuda-toolkit-10-1
  • set PATH variable if necessary
  • have nvidia-persistenced start on boot and make sure it is continuously running
    done.

Edit: if the normal driver package tries to pull in Xorg, please use the “-headless” package instead.

Where are the tools? I searched for nvcc and they are no where to be found… If I do a normal install with cuda_10.1.243_418.87.00_linux.run I get all the tools but I don’t see these with the steps above.

The .deb installs the binaries to /usr/local/cuda-10.1/bin/
Please see this about the location of the files and the necessary settings:
https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#post-installation-actions