The current implementation of nanite and raytracing cannot really exist together in practice and so it uses a proxy mesh for ray intersection but is that the only solution for combining raytracing and nanite? I think when it comes to using unreal for more cinematic purposes, raytracing is possibly one of the most important feature sets for achieving great visuals and it would be a shame to lose out on it when using nanite. While virtual shadows are great they still are not on par with rt shadows but maybe we could have a hybrid where virtual shadows are only used for nanite meshes meanwhile rt shadows can be used on the rest of the meshes if it be possible.
In all likelihood, we’ll only ever see proxy meshes.
Reason being: the way hardware raytracing handles triangles requires a lot of memory, a lot of a lot of memory, quite a bit worse even than normal geometry. To the point where Xbox Series S doesn’t support RT on a lot of games coming over from last gen sheerly out of a lack of RAM, so there’s little chance, and no way I can think of, that any retail machine is going to have enough memory to store hundreds of thousand or millions of triangles per asset for hardware RT.
But then this probably won’t be the end of the world. The non nanite geometry pipeline could benefit from having a high priority for updating, without going to a full “nanite” like system, rather early just to have it work with Lumen and virtual shadow maps better, as that’s going to be a potential issue for any game that wants to ship. Some sort of “hybrid” model between VSM and RT shadows isn’t impossible, but I can’t imagine whether that’s going to be a priority for Epic over say, the above. Especially with interactive foliage being very difficult with RT shadows and being non nanite as it is.
Very true unfortunately, I hope at least that virtual shadow maps become a lot better as the technology evolves.