I have 2 computers - a laptop and a desktop. Both have somewhat similar specs (16 gb ram vs 12 gb, both running SSDs, both having decent processors - i5-2500k vs i7-4700hq, 4 vs 8 threads). The slower one is the desktop.
However when it comes to building lighting, one takes 4 hours whereas the other takes 5 minutes. I understand that specs could make a smaller difference in this, but I doubt it’s a difference that makes it roughly 50 times slower. I’m struggling for a solution here. I’ve tried flushing the swarm cache folder, cleaned and validated it to no avail. Another difference is that the 4 hour compile gets stuck around some low % in Swarm. Sometimes it’s 5%, sometimes 7%, but always below 10%. Swarm keeps building however; it is not unresponsive and eventually finishes.
I was wondering if anyone has had similar problems and/or know how to solve an issue such as this?
The reason this is happening is because of the size of your lightmaps. Check to see if you have objects that have very high lightmap resolution. I’m no expert on this but, the way the load is distributed is across the computers is, each object’s lightmap is computed by one thread of the PC. So if you have 100 objects that have say 64 lightmap resolution, and have one object that has 1536 lightmap resolution, those 100 objects would finish up very quickly on one PC and the object with 1536 resolution would go for 4h, on the other PC.
The way to fix this is to make all the objects have similar lightmap resolution. If you have a big object that requires a bigger lightmap res, then simply divide it into smaller objects and reduce the lightmap resolution.
Hi, thanks for replying. You seem to be mistaken in what I was looking for, however. I didn’t want to network two different machines, I only used the secondary machine to test out if the result was the same.
Besides that, my lightmap values are low, and it wouldn’t correspond at all since one machine build the lighting in a timeframe of 5 minutes whereas the other took several (4+) hours.
I found out on my own what was causing this - a material that got corrupted by moving from 4.12 to 4.13 because of the new geometry fix in 4.13. Recompiling this material fixed this and another issue I was having.
