Hi Prehabitat,prehabitat wrote:This thread is heading towards adaptive sampling right? Isn't that already locked in as an octane 3.1 feature?
Thanks for the tip regarding adaptive sampling, I did not know about this feature or how it is done.
After having fun yesterday analysing how octane engine is handling noise and how long it takes to "converge" to a clean image, It is clear that the engine has a good sampling strategy. There seems to be a compromise though to keep noise pattern fixed for animation, but it is a reasonable compromise although, in my mind, it may slow down convergence for stills. I could be wrong.
I let Octane render the scene to 50000 spp to see how clean a scene could get if you go extreme and obtain a reference image (attached). the scene pixels RGB channels, however, already converge to almost final values by 20000 SPP. I attached a typical example of RGB channels updating for a single pixel. of course, this is specific to this scene but it gives you a general idea of performance.
I have also checked if, by knowing the updating trend of lower SPP values we can extrapolate to final converged values. my conclusion is that it can not be done due to the way the engine update the samples leading to RGB values jumping around without clear trend.
It was fun to explore the subject, however, and see Octane in action.
Regards,
Mishal
