But if you do a quick render, estimate how long each sample takes, then decide how long you can stomach waiting, and set the denoise to start close to there, the end result is a faster render while minimizing lost quality.
I tested this many times. I always run many tests for these sorts of things to see what happens and there was no improved quality if you started denoising later on, nor was it any faster. I suggest you run a few tests yourself to see. As a matter of fact, I had a render go FASTER with denoising on, which honestly didn't make sense to me, but I can't argue against what I seen myself. Plus I posted my results in here, one with denoising that took 5 mins, one without that took 45 mins and they look identical.
I have since run yet even MORE tests lately and in SOME of my tests I can see the difference, but you need a magnifying glass and several minutes to see it and most of the time, the difference is more noise from the render that didn't have denoising... and it looked worse. In any event, if you're rendering for something like a visual novel, the odds are t he person who sees it will probably see a compressed image like a WEBP or a JPEG, and they won't sit there for 10 mins examining the pixels, but will probably zip by it fairly quickly.
What can I say, I ran many tests. Most I could not see a difference, the denoising never took longer for me, not with my older crappy video card and certainly not now with my newer card. Certain renders are more noticeable than others, but in the tests where it was, the difference wasn't worth the additional time required.
This topic reminds me of NTR, people seem to get their panties in a bunch over the mere suggestion of it, well, I may start adding people to my BLOCKED list on here so I no longer see their whining (and unless you offer an alternative solution when you complain, than it is merely whining). I'm getting tired of discussing it. I offer advice based on my own tests, don't believe me or want to argue against it, fine, whatever, all I can say is to those considering it... RUN YOUR OWN TESTS AND SEE FOR YOURSELF! All I know is what works for me, and I won't stop using it. Period.
Anyhow... back on topic... another horribly denoised image by me... avert your eyes people!!!