Thought this might be useful: both images rendered in IRay for 550 iterations, both oversampled (rendered at double size then reduced to normal), but image 2 uses the built-in post denoiser (with "Post Denoiser Start Iteration" set to 550, the last iteration).
Not perfect, but handy for clearing up shortish renders if you don't like the grainy look.
View attachment 1883896
View attachment 1883897
This is exactly the point I was trying to make, lower resolution and lower converged images can benefit from denoising, but to say it looks exactly the same is just simply inaccurate. While this is a good test, it isn't perfect. Your artificial limit of iterations is a compromise. Given two images that are only converged to the same percentage, denoised will almost always look better. Now that, in the real world, is not how or why we use denoising. In general the reason to use the denoiser is to lower the render times. We substitute denoising for higher image convergence. That is exactly what most people do, myself included when I need something fast. Nobody wants their PCs to run for hours on end just to get one image. But if you compare an image using DAZ default render settings and not limit the iterations to 550 then compare that image to a limited 550 iterations / denoised image you would see that the DAZ default will have resolved more detail. By all means, if you are always going to limit the iterations, for lack of hardware or lack of time or whatever reason, please use denoising, your images and all of our eyes will thank you for it.
But when denoiser fans say it is identical, they forget the other factors that made them use a denoiser in the first place. I ran a similar test to you with the following difference, I did not limit the normal image (not denoised) to the same amount of iterations as the limited version. There is a very visual difference in the level of detail. Individual hairs can be seen in the non denoised image and they are simply blurred in the denoised image. The same blurring can be seen in the shadow areas. In both these cases, on my setup, there is an obvious difference in render times, but we are not talking about minutes versus hours, it is more like minutes versus tens of minutes. A small price for me to pay for a more detailed image.
Does this make any bit of difference, not really. I don't think I have ever seen any VN or Comic here on F95 that has a base resolution higher than 1080p, nothing at 4K or higher. Almost all use single images, which can take advantage of denoising. As a matter of fact, I would actually like if more people that post here would use denoising. There are a shit ton of noisy images, and some spectacularly posed and lit renders look horrible with the amount of noise in the images.
I am not anti-denoise, never have been, just realistic about what it does and when to use it. Generally, I don't use denoising. In my situation I am not trying to pump out tens or hundreds of images a day, and I have enough hardware where I can still do my CAD work while an image renders in the background. I also render oversize and to a very high convergence with high iterations and high time limits, then scale the image back down to about 1440p. There are a lot of benefits for some to use denoising and please continue using it, just don't misrepresent what it can and can't do. Is that too much to ask? (rhetorical question to be answered only for yourself).