Hi Alex,
Thanks for your answer. I’ve got a quick overview from our Team.
Please note: not all of this has been deeply investigated on our end yet, this is just a quick list of initial thoughts on our end.
GPU LM:
+ Should be theoretically faster to bake.
+ Has the community branch which seems fairly active.
+ Better support of emissive materials.
- Inconsistent results
- Reflection Probes don’t always bake properly
- Does not handle multiple lighting scenarios properly
- Lacks controllable settings.
- Doesn’t work well with automation.
- Crashes without information when map check issues are present.
CPU LM:
+ A lot more granular control over settings.
+ Handles multiple lighting scenarios properly.
+ Swarm
+ Works well with automation
- Slower on a single machine.
- May be missing certain features.
- Rect lights don’t seem to work as well on CPU Lightmass.
We’ve generally seen a lot of inconsistencies between the two methods. But GPU Lightmass seems very unreliable, from what I’ve gathered from our team.
Thanks,
Aidan.
[Attachment Removed]