I currently have a GTX 1080 ti that I use as my main GPU, and also for CUDA workloads. I was wondering if it would be possible to use an AMD GPU as my main GPU, for things like gaming, but keep my 1080 ti as a dedicated card for CUDA compute (and preferably things like PhysX and RTX Voice). Or perhaps would it be more appropriate to use a card intended as a pure compute card? Or would it not work at all?
I’m not sure if there would be any compatibility or driver issues, and I’m using Windows 10 (since that probably matters).
Thanks for any help, I’m pretty new to CUDA
Why not use two NVIDIA GPUs? NVIDIA offers a great variety of products at a diversity of price points and performance levels. I don’t game, but over the past decade have often used a combination of a low-end Quadro to service the GUI and a high-end GPU (Tesla or Quadro) for compute.
Anecdotally, I have read reports of people who claim to have successfully run an AMD card along with an NVIDIA card in the same Windows DX12 system. Personally, I wouldn’t risk it. If there are any issues, each company’s support would likely point to the other company’s components as the source of the problem.
I may end up doing that, just wondering if there was in any way the option to mix, if (for my use case) it may be benefitial. But, yeah, the risk definitely doesn’t seem worth it. If I ended up jumping outside of Nvidia, I’ll just move my GPU to another PC or sell it and pick up one of the Jetson line-up, so I can continue learning CUDA (and force myself into using Linux). Thanks.