3 ms·

It’s all path traced (with denoising and DLSS upsampling I think). The poly count is much less impressive than the number of different materials and number of

It’s all path traced (with denoising and DLSS upsampling I think). The poly count is much less impressive than the number of different materials and number of light sources.

brundolf 6y ago
> The poly count is much less impressive than the number of different materials and number of light sources. Is it? The basic tracing algorithms I'm familiar with scale linearly with the number of polygons (minus whatever can be culled out), and don't really care about the number of materials (other than for memory usage I guess) or light sources (because the rays come from the camera). I do know that for edge cases like refraction some more advanced algorithms will do partial forward-tracing from the light sources, so maybe that's what you're referring to. Though I didn't notice any refraction in this demo (it wasn't a big focus, at least).
- pixel_fcker 6y ago
  All raytracing algorithms are based on some spatial subdivision structure (usually BVH based) which gets you O(logN) instead of O(N). Increasing poly count only gets interesting when you figure out how to compress truly huge scenes into a limited memory footprint. Lots of different materials are difficult for GPUs because of divergence - if every ray spawned from a surface scattering event hits a different material you’ve lost all parallelism. Lots of lights: Evaluating them all for a given surface scattering event is obviously an O(N) problem (for 10 lights no big deal, but what about thousands?), so you want to choose some small subset of “important” lights to consider. Making that choice is a hard problem because the correct choice depends on the product of the incident radiance from the light, the BSDF at the surface as well as the visibility function. There’s a lot of interesting research being done on this (Path guiding by learning the incident light field or its product with the bsdf, new sampling techniques, spatio-directional acceleration structures) at the moment.
  - brundolf 6y ago
    Got it, that makes sense

It’s all path traced (with denoising and DLSS upsampling I think). The poly count is much less impressive than the number of different materials and number of light sources.

brundolf 6y ago

> The poly count is much less impressive than the number of different materials and number of light sources. Is it? The basic tracing algorithms I'm familiar with scale linearly with the number of polygons (minus whatever can be culled out), and don't really care about the number of materials (other than for memory usage I guess) or light sources (because the rays come from the camera). I do know that for edge cases like refraction some more advanced algorithms will do partial forward-tracing from the light sources, so maybe that's what you're referring to. Though I didn't notice any refraction in this demo (it wasn't a big focus, at least).

pixel_fcker 6y ago
All raytracing algorithms are based on some spatial subdivision structure (usually BVH based) which gets you O(logN) instead of O(N). Increasing poly count only gets interesting when you figure out how to compress truly huge scenes into a limited memory footprint. Lots of different materials are difficult for GPUs because of divergence - if every ray spawned from a surface scattering event hits a different material you’ve lost all parallelism. Lots of lights: Evaluating them all for a given surface scattering event is obviously an O(N) problem (for 10 lights no big deal, but what about thousands?), so you want to choose some small subset of “important” lights to consider. Making that choice is a hard problem because the correct choice depends on the product of the incident radiance from the light, the BSDF at the surface as well as the visibility function. There’s a lot of interesting research being done on this (Path guiding by learning the incident light field or its product with the bsdf, new sampling techniques, spatio-directional acceleration structures) at the moment.
- brundolf 6y ago
  Got it, that makes sense