Get Huge SDXL Inference Speed Boost With Disabling Shared VRAM – Tested With 8 GB VRAM GPU