I am just putting this out here…
Firstly, I am very happy with Cycles. I think it is a great addition to Rhino 3D.
Secondly, while it might not be best for final renders, nVidia’s denoise works really well to eliminate rendering noise.
While talking with a friend about the placebo-limit to Cycles rendering with a friend, which seems about about 12,000-15,000, passes, I was reminded that the noise was more often light noise in dark areas which are noticeable, such as this 1700-pass render:
It just may be me, but I don’t notice the dark noise as much in the light areas, and i don’t think it’s a perception thing, and even if it was it could be work on like Bayer Arrays in digital sensors.
Whilst, nVidia’s Denosier is very-much optimized for path-tracing, it might be possible to get a sharper image yet, by exploiting the bias to have light-noise instead of dark-noise.
In other words, the denoiser is probably unbiased as to it fixes light noise or dark noise. Perhaps it doesn’t need to fix dark noise, or as much, resulting in a sharper image.
Perhaps, some experimentation could be done in PhotoShop, with layer modes to test it. Hmmm.