...
Graphics cards that support CUDA platform (Nvidia), running on Windows. No AMD or Intel will work, sorry. This means the GPU engine is not supported for Mac, as they don’t use Nvidia cards.
The drivers should be up to date.
Maxwell, Pascal, Volta, Turing and Ampere architectures are supported (for reference you can check this table); it could work on Kepler architecture but it could be unstable and the experience won't be good. Currently, the graphics cards using the Ada Lovelace architecture are not supported.
Denoiser
Denoiser is capable of running on CPU or GPU, with Nvidia, AMD and Intel graphics cards; it works with CUDA (Nvidia) and also can work using OpenCL either in GPU (Nvidia, AMD or Intel) or CPU.
When using GPU, we recommend having a graphics card with at least 2.5GB. This number is actually dependent on frame size, 2.5GB is for 2k pixels square imagery.
When using Nvidia CUDA, the drivers should support at least CUDA 7.0
When using AMD, the Catalyst driver should be up to date.
In the case you get an error message when using CPU about OpenCL ICDs, it will probably be solved by installing one of these packages depending on your OS:
Windows: opencl_runtime_18.1_x64_setup.msi
The Denoiser will first try to use CUDA, if it doesn't find a compatible card, it will fallback to OpenCL GPU (usually with AMD and Intel cards). If the image doesn't fit in the graphics card memory, you can then use OpenCL with CPU (which will use RAM memory).
...