CUDA Installation Guide for Ubuntu

- 1 min

Anyone doing machine learning right now is probably familiar with Nvidia’s GPU computation library CUDA. It delivers an incredible performance boost that makes training deep neural networks possible on even a single workstation.

However, for all it’s awesome performance CUDA can be hard to install and sometimes interferes with the graphics drivers and the display manager.

Therefore I wrote this small installation guide in case I would ever need to reinstall my computer. It is based mostly off the official Nvidia guide but I gathered all the information in one place with a section containing fixes to problems I encountered.

I’ll try to keep it updated but for now, I think the method should work for the recent versions of Ubuntu (>= 14.04) and CUDA (>=7.5).

The guide is hosted as a Gist on Github, but I also embedded the guide below for convenience.


Erik Gärtner

Erik Gärtner

Research Scientist at RADiCAL