by frankmci » Mon Oct 18, 2021 10:20 pm
frankmci
Mon Oct 18, 2021 10:20 pm
In my experience, denoising speed is not directly linked to render speed. A 1:00 per frame render might take 0:13 to denoise while a 0:07 rendered frame of the same shot might take 0:07 to denoise. It's not strictly proportional. Once you get down to single digits for render times, it may be faster to increase sampling and disable denoising to achieve similar image quality. For me, my rough rule of thumb is if each frame is going to take more than 20 or 30 seconds to render with acceptable quality, then it's worth seeing what the denoiser can do for me. I assume it will always add several seconds, so it obviously needs to shave off significantly more time than it adds to be worthwhile. When each frame is a minute or more, it's usually worth using, although there are some lighting situations where denoising is still likely to produce some noticeable artifacting, so we just bite the bullet.
In short, the behavior you are describing sounds normal to me.
I should say that our shop is currently locked into Octane 2020.2.5, and probably will be for six months to a year, so I haven't experimented much with newer builds.
Technical Director - C4D, Maya, AE, - Washington DC