Hello,
I had this issue today. I have a Lenovo P53 Thinkpad with an RTX GPU. I was in Blender 3.1 and noticed for the first time that I didn’t have the option to choose my video card. Blender claimed not to see my card. I have been doing a lot of NVidia driver and library installs lately to get some ML scripts to compile and realized that the environment variables that were new to my system might have something to do with it, so I renamed some of the variables by putting “TEMP_” in front of each of these and Blender now sees the GPU. Examples below with the values:
TEMP_CUDA_PATH
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.6
TEMP_CUDA_PATH_V11_6
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.6
TEMP_CUDA_VISIBLE_DEVICES
1
Blender began seeing my card again. The CUDA_PATH entries don’t have anything to do with it, so I set those back, but when I put the CUDA_VISIBLE_DEVICES variable back as original, it couldn’t see my GPU. So I changed the value to 0 and it sees it again. I have an integrated graphics card (Intel) on my machine also, which it’s also seeing when set to 0, so either renaming the variable or setting it to another number based on your situation may help.
When CUDA_VISIBLE_DEVICES is set to 1
“No compatible GPUs found for Cycles
Requires NVIDIA GPU with compute capability 3.0”
When CUDA_VISIBLE_DEVICES is set to 0
I hope this helps, because I looked at a few sites for a solution and didn’t see an immediate fix except for reinstalling the drivers, which wouldn’t have worked for me.
Thanks for an amazing community and application.
-EsDog