Logo

Indie Machine Learning and Video Game Dev

Neural Net Downloads
How to
Hardware Reviews
Get Merch
About
Training of the week
5 August 2019

Mixing AMD and NVIDIA gpus on same system

by Mathieu Poliquin

amd_nvdia

I was curious to see one can mix AMD and NVIDIA cards and make them run Tensorflow at the same time on one system. The advantage, for my case, is that it avoids having to maintain a seperate system to test AMD cards.

Conclusion: It works! and quite well actually!

How to install

1. Install CUDA 10 + tensorflow-gpu

Just install it as usual CUDA Install guide

pip3 install tensorflow-gpu

2. Install ROCM + tensorflow-rocm

ROCM install guide

The only difference is that you will need to install tensorflow-rocm in it’s own virtual env so it doesn’t conflict with vanilla tensorflow

Install virtual env

sudo pip3 install -U virtualenv

Create a virtual env called “venv” in your home directory

virtualenv --system-site-packages -p python3 ~/venv

Actitvate the environement before installing or using tensorflow-rocm

source ~/venv/bin/activate

You should now see “(venv) $” at the beginning of your shell prompt

Install tensorflow-rocm

pip3 install tensorflow-rocm==1.13.4

note that I specified 1.13.4 because 1.14 crashes for me

When you are finished using that virtual env you can quit with this command

deactivate

Details

Software

Hardware

amd_nvdia_hardware

A video of the test:

tags: amd - nvidia - gpu - tensorflow - machine learning