Current main computer on which I run ML:

• AMD 3700X / 32GB RAM / 6700XT 12GB
• Windows 10 / Python 3.10
• Stable Diffusion (ONNX via DirectML) 1.5 and 2.0

Everything is done directly from code, no UI. A CUDA card would probably be faster, but I’m very hopeful AMD is going to solve this and provide better tools to run models on Windows.