Hey all,
I've been testing out the XB0 release of 2022.1 for some caustics renders and have a specific question.
I know this feature is very early development but I hope by feedback is received well
So far I've managed to produce an image with sidefx's Mantra engine that I feel represents the most "accurate" example I've come across and I've been trying to match it in other engines to hopefully resolve the image faster. Redshift was giving me fairly OK results but when I pushed the photon count to the max my RTX 3090 could allow it still came out more noisy than Mantra and was slower to render. This is where Octane Photon tracing kernel seems to shin! The memory overhead is the lowest of any engine I've tested and there is so waiting for photon map generation, it just starts instantly and resolves quickly, nice!
So with the good news out of the way here is the bad; there is either a missing indirect component or I can't find where to "enable" this in the software.
First here is my "master" reference created with Mantra, you can see there is some light scattering within the prism so its not completely black and there is indirect light bounce back onto the ground plane as well. To achieve this in Mantra I'm using 2 indirect lights "Caustic Photon Map" + "Indirect Global Photon Map" with the PBR render settings on "All Paths" vs the default "Specular, Diffuse Paths". Render time about 55s
The redshift result isn't bad, but prism is very dark, the bounce light onto the ground plane is working well though, here there is more noise than the mantra render and the render time was slightly higher at about 70s
Now onto Octane, this first result is with bounce settings matching Mantra, diffuse depth of 2 and specular of 6, at 2000 samples it rendered in 19s. But the scattering within the prism is missing and there is no diffuse bounce on the ground plane.
I tried increasing all the bounce settings along with a 10x increase to the light strength and you can see there are photons bouncing around there but after 10 minuets rendering this is the result I got, its very noisy and still far from my reference
Can anyone point me to either where I'm missing something or a dev chime in with why there is this difference? I'm happy to provide orbx file if needed!