Here is what mega geometry does. Basically it collects triangles together into clustered regions where it makes sense so that there are fewer collision calculations necessary for more dense geometry-ray interactions, which are becoming the default lighting systems in games (even Lumen, a hybrid system, has them.) Again, it doesn't use AI/neural networks/machine learning at all. With it off, you have to have fewer, often larger, triangles for the same compute, which will look worse in-game.








