Currently Online 
Cerebralbore101 said:
The bolded makes absolutely no sense whatsoever. Just sounds like gobbledygook. What's non-ML? What's a sub-mesh? What is BVH, BLAS, and TLAS? So mega geometry is just something to make ray tracing easier? Raytracing is a waste of resources IMO. I've never been impressed with it. |
non-ML - not machine learning, which is the majority of what we call "AI" today.
BVH - bounding volume hierarchy (a class of data structure, ways to structure data, that help ray tracing algorithms identify geometry-ray boundaries, efficiently. They are used for other use cases though too, like physics or certain forms of culling [removing objects from a scene].)
BLAS and TLAS are just types of BVH's that work at different scales. BLAS works at the level of individual objects and builds up, TLAS works at the level of the whole scene and builds down.
A mesh is the skeleton of a 3D object a sub-mesh is a part of that skeleton.
Mega Geometry just adds an intermediate data-structure where groups of triangles are treated alike because they are nearby. It makes real-time Ray Tracing easier, yes, by allowing it to compute on less geometry to achieve the same result.
Ray tracing is the present and future of real-time rendering, and has been a staple of CGI rendering for decades.
Neural shaders scale according to hardware, so no not just low-end. Also compute heavy =/= high quality. The whole point is that neural shaders can be scaled to higher qualities than non-neural assisted shaders, because you can have more of them, more intricate shaders, etc as you save on the per shader compute. The end result usually is higher quality or more performance at the same quality. Because it is trained on a per-game basis, it is perfectly compatible with the artistic-intent.
Last edited by sc94597 - on 10 May 2026
