Hi! Please help. ComfyUI is not using GPU1 (RTX 3080 Ti Laptop) every now and then, it uses GPU0 (Intel Iris Xe) and CPU instead. This usually happens at the first generation with a new prompt even though the model (SDXL with refiner) is already loaded. All the next generations run fast, but with the slightest change in the prompt it begins with an at least 10x slower generation. What could cause the problem? Any help is much appreciated.
1 Like