We frequently see tutorials about how to speed up your render and I want to be fair, they are mostly nice.
Although, most of them, or all I have seen tell us about using denoiser.
WOW, nice!
Sadly, I did not find any who made it clear “USE DENOISER IN THE COMPOSITOR”, cause we still see so many people that refuse to use compositor for N reasons and prefer to use the denoiser during the render process.
Why is that wrong?
Well, I have seen many comments about the fact that the denoiser in compositor gives better results. I would say they are a bit better results really but not just that. IT’S MUCH FASTER TOO!
I always wanted to understand why people always complain about render times, make huge efforts to reduce render times for a few seconds but at the end they render with a denoise system that is slower and gives worse results. It’s like taking water from a boat sinking and throwing the water inside the boat.
One of the reasons it’s faster is because when you use the denoiser in the render Blender uses just half of your CPU threads to compute the render. Did you see that? All the rest is destined for the denoising pass… and it needs to wait for the other threads to work arround a complete thread area to start computing the denoise. AND IT’S SLOW! Don’t ask me why.
BUT when you use the denoiser in the compositor all the threads of your PCU are used to help render the image. I believe it’s not necessary to prove that a hole set of threads is faster that half of them, right?
I have no idea how the algorithm works on the compositor denoiser but believe me, its much faster than just wait for the half of your processor to denoise in render.
And one example. The first image is rendered using denoiser in the render and the second one using compositor denoiser. Of course, both are open image denoiser.
So, if you care about your render speed, just stop using denoiser in the render and use in the compositor.
What do you think you could do in this 40 seconds you spare having a better render?