Low GPU usage

I've currently set up a pod with 4 different GPUs and allocated each GPU to a different port with the command: CUDA_VISIBLE_DEVICES=0 python main.py --listen --port 8188. It worked. I have 4 different ComfyUI tabs operating by themselves. But when I generate, the speed of generations are incredibly slow. It has taken 2+ minutes to generate a single SDXL image on all of them.