… between precomputed realtime GI and baked lighting given a scene with:
I know that I can run a test scene and look at the profiler, however I am curious about the theory here: I would have thought that the performance would be the same. Is that the case? And if not, then how come?
Thanks
Assuming by baked you mean using “Baked indirect” (so the directional light still applies realtime direct light), performance should be similar. I think someone found that precomputed was faster in their use case, but I guess it all depends on the precomputed resolution and the baked resolution.
it depend at what kind of performance you looking at.
looking only at framerate, baked lighting will probably be much cheaper because it will not need to render shadow for static objects, causing less drawcall, but it will increase memory usage by making lot of lightmap textures.
and if you have several lights they will be free because it will not increase the amount of lightmap, they just add up in the same texture.
If you bake Gi with real time Gi, it still have to draw all shadow at runtime, increasing amount of drawcall, and if you add additional light they will have to be render one after another with thier shadow as well. (its cheaper to have lots of light in deffered rendering though)
real time lightmap also use memory to store global illumination, but not as a texture, and its usualy less heavy than lightmaps
also real time Gi have a process that use a relatively small amount of cpu time to update the Gi when lighting condition changes at runtime.
Thanks for the replies - so if shadows were disabled completely and we are considering only FPS they should be pretty similar right? I have outdoor toon-style scenes and I’m considering just disabling shadows and using real-time GI to side-step baked lighting issues. So far the performance is looking the same but I thought it’s too good to be true and I must be missing something. I guess not!