As I attempt to teach myself as much as I can using the internet, I sometimes encounter contradictory things, or assertions that don’t seem to make sense, so I’d like to ask the Blender hive mind.
I’m rendering with Cycles, rather than doing anything with game engines, which is what the discussions of this sort of thing seem to usually talk about. I’m also trying to make good use of instancing, and texture atlases. But the question is about how to group objects for atlasing.
Obviously, for a given texel density I could unwrap a collection of objects into one atlas with a shared 4k texture map set, or I could unwrap them into four different texture atlases with four separate 1k texture map sets. Assuming that I don’t have awkward shapes that make the packing density different, it SEEMS as if these ought to amount to the same thing in terms of memory usage and performance, at least with raytracing.
Am I missing something? Is there a reason why, outside of games, you’d want to organize things one way or the other?