I'm curious about this kind of thing from an engine and console architecture perspective. Any gamedevs able to shed some light?
I work in the industry, but not directly on low-level engine implementation details. Personally, my gut thinking is that the Creation Engine is falling behind in terms of modern asset streaming techniques.
In an imaginary world where I've poured over Bethesda's engine source for days, I wonder if I might discover that:
-
Asset formats and/or orchestration code used for asset streaming in the Creation Engine are not optimized to a degree where scene graphs can be effectively culled based on camera frustum or player proximity without noticeable dips in frame-time. It simply takes too long to pause actor simulations or too long to stream assets back into memory and reintroduce objects to the scene graph
-
Virtualized geometry or other magical low-overhead auto-LOD solutions aren't in place. As far as I understand it, efforts like Nanite in UE5 were an enormous engineering investment for Epic, and unless Bethesda has a massive engine team of their own (they don't), they simply won't be able to benefit from an in-house equivalent in tech