[SOLVED] GPU computing platform for AMD architecture

Issue

As we all know trying to train a deep neural network or basically any type of AI model requires a significant amount of computation power. Doing this on CPU can take up to days depending on the project.

I have a AMD GPU specisifically the model name of GPU is "RX5700". For NVIDIA GPU’s there is CUDA which can easily be used for GPU computing however, I cannot find an alternative for AMD architecture. I know there is ROCm but that only works for very special GPU series and Ubuntu operating system.

Is there any platform where one can do GPU computing on Windows with RX series GPUs?

Thanks for the answers.

Solution

When I had a Radeon 6700XT I was able to leverage Microsoft’s DirectML. It wasn’t as good as NVIDIA/CUDA/TensorFlow etc. Although it was still much better than training on a CPU. It was quite limiting though as it was still using a much older version of TensorFlow.

Answered By – Ace

Answer Checked By – Jay B. (BugsFixing Admin)

Leave a Reply

Your email address will not be published. Required fields are marked *