Hi, My PC has two graphics cards, both NVIDIA. One for exclusively running render jobs (Blender) and one for the desktop, so that the desktop does not get slowed down when running blender renderings.
That has worked fine...until now...(cannot remember exactly when this happes since I do not render each day...) To get that working I set the CUDA device to be used inside Blenders User Preferences to one graphics card and set the other one via the environment variable CUDA_VISIBLE_DEVICES for X and the desktop. Has something changed inside the nvidia drivers, which prevents this setup to work? Is there something known which relates to this problem? (I searched the nvidia drivers docs on my harddisc and found nothing and the internet gives me to much too old hits related to CUDA_VISIBLE_DEVICES to be useful...) Cheers! Meino
