CUDA Failure - Troubleshooting

[SamuelAB]

One of our R&D engineer also made an algorithm to split meshes automatically into equal chunks super quickly, so that may come in handy.

[karl]

Another thing you could try is disabling RTX acceleration in Octane preferences, since there will be some data we don't need to prepare in that case.

[jimho]

to chop the big trunks into small pieces (inside octane automatically) will be important for Octane,
not only for meshes but also for big textures(e.g. 8k map), as I remember this is the bottle neck which also make Nvlink not working well...
hope this issue can be resolve soon.
viewtopic.php?f=33&t=73978&start=30#p378155

[SamuelAB]

By the way, I do high quality VR art pieces and as I mentioned, for video-games (especially VR), you want to have as few objects as possible. So that means you will have few objects, but they are all huge. I also bake all my lighting with Octane because it is so fast and beautiful.

So while I will be able to split my scene in this case because I'm not making a VR game, it would be good to find a long term solution to this that allows users to load giant scenes with large chunks of geometry (to reduce draw calls)so that Octane remains a good solution to bake the lighting in these scenes