Post #3201625

Current setup of multimodel programming is limited by the amount of VRAM I have to share, but what about getting rid of this limit? I just got an idea while commuting to work lol.