Tutorials:Install nVidia CUDA and GPU drivers

Overview

This tutorial will walk you through the necessary steps to set up nVidia CUDA and the correct corresponding version of GPU drivers, as well as a bunch of libraries needed for development. Note, before you do anything else, you should first update your system packages to the latest versions,

#!/bin/bash
sudo apt-get update
sudo apt-get upgrade

Guide

If you have a freshly installed Ubuntu 18.04 or some derivative (such as Linux Mint 19), this should be simply a matter of running the following script (run with sudo):

#!/bin/bash
wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-repo-ubuntu1804_10.1.168-1_amd64.deb
dpkg -i cuda-repo-ubuntu1804_10.1.168-1_amd64.deb
apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub
apt-get update
apt-get install -y cuda

In case you had the driver or an earlier release of CUDA already installed, my recommendation is to first uninstall them.

If you cannot get the script to run or have a different linux version, head over to the nVidia CUDA download page and install CUDA 10.1 by following their instructions. I strongy recommend the networked ".deb" package installation. Make sure to explain any problems you encounter on the discussion page so that we can troubleshoot this guide.

Post installation, you should reboot your system to make sure it still works. Run the following commands in a shell to test installation, which should produce a similar output than what is shown:

# show version of the nVidia driver
> glxinfo | grep NVIDIA
TODO: copy reference output

# show version of the CUDA compiler toolchain
> /usr/local/cuda/nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Wed_Apr_24_19:10:27_PDT_2019
Cuda compilation tools, release 10.1, V10.1.168

# show detected GPUs
> nvidia-smi
Sat May 18 12:51:46 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.39       Driver Version: 418.39       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Quadro GV100        On   | 00000000:03:00.0  On |                  Off |
| 35%   48C    P0    27W / 250W |   1490MiB / 32475MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      1354      G   /usr/lib/xorg/Xorg                           800MiB |
|    0      1762      G   cinnamon                                     331MiB |
|    0     10767      G   /usr/lib/firefox/firefox                       6MiB |
| <... more processes>                                                        |
|=============================================================================|

You should also add the path for the binaries and libraries to your environment so that all tools are able to find them. For example, add the following two lines to your ".bashrc" in your home folder:

export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH