Comfyui and GPU on different machines?
retained text
Can I run comfyui on one machine (no GPU) and point it to llama-swap and/or stable-diffusion.cpp (or something similar) on a different machine (with GPU)? If so, what does that stack/setup look like? I've been googling this all day and haven't really made any progress. Not looking for suggestions like "just install comfyui on the second machine and access it from the first" - I'm specifically trying to achieve a certain level of service segmentation and modularity in my setup.
document fetched 2026-10-11 00:37 · canonical https://reddit.com/r/comfyui/comments/1x2nt75/comfyui_and_gpu_on_different_machines