Seeking Setup and Configuration Advice
retained text
To frame my question: I'm experienced creating AI images with hosted models. For instance, writing effective prompts, using seeds, and reference images. And get good results. But I have no experience with self-hosted models and related tools. Things like ComfyUI, LoRAs, and custom workflows. Here's a description of my desktop computer: \> Mac Studio 2025 w/ Apple M3 Ultra with 60-core GPU, 96 GB memory, and Apple SSD 2 TB drive. External Drive: SanDisk G-DRIVE PRO - Enterprise-Class 12 TB This may seem excessive, but I work with large datasets using memory and processor-intensive mathematical and statistical models. The reason for noting my computer is about compatibility. A friend using a Mac can be a problem for these reasons: The hardware and memory are non-issues. But there is no CUDA. ComfyUI on Apple Silicon runs on PyTorch's MPS/Metal backend. That means the entire NVIDIA optimization stack is simply absent. Any tutorial you follow will assume you have it. Also, fp8 checkpoints aren’t supported, so common "fp8 scaled" FLUX and SD3.5 files will error out. To be clear, I don’t know if he's right or wrong, which is why I'm asking. If you think a web-based host is best, I'm open to suggestions. I appreciate any advice you can offer.
document fetched 2026-10-25 00:36 · canonical https://reddit.com/r/comfyui/comments/1xduqdb/seeking_setup_and_configuration_advice