I use a headless Xubuntu Server 18.04.3 with two GTX 970 cards as a render farm. This has worked with Blender up to 2.79 but is now broken in 2.80.

Basic command to that used to work:

blender -b foo.blend -o foo_ -E CYCLES -F AVIJPEG -P cuda0m.py -x 1 -a

where cuda0m.py was:

import bpy, _cycles

bpy.context.scene.cycles.device = 'GPU'
prefs = bpy.context.user_preferences.addons['cycles'].preferences
prefs.compute_device_type = 'CUDA'

I fixed one API error (user_preferences is now preferences in 2.8) and the render runs but not on the cards. I'm thinking I'm not pointing Blender to use the GPUs correctly.

If the python is correct, it could be another issue: I installed 2.80 with:

apt remove blender
snap install blender --classic

...and I'm seeing a warning line when I run the script that could conceivably be the culprit:

/run/user/1000/snap.blender/gvfs/ non-existent directory

Driver version: NVIDIA-Linux-x86_64-410.66.run


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.