Yep understandable 
Just out if interest, did you guys finally managed to get altus on board and is the denoiser part of that partnership or did you guys started from scratch?
Well, as you know, originally I didn’t intend to write our own denoiser and I said so many times. I was hoping that the guys at Innobright will be able to fill that gap and we were talking with them about a tighter integration. However at some point it became clear that they will need more time to get their stuff together, and we didn’t really have that time - so we had to roll up our collective sleeves and do it from scratch. I had to sift through any papers that I could find and test out different approaches with my colleagues until we figured out something that worked well enough. Also, our GPU team rose to occasion to produce a GPU-accelerated version which so far seems to work very well.
Best regards,
Vlado
Hear Hear!
Right, well, well done guys, great tool and it really seems awesome
Big THANKS for Vlado and the team to pull this off to implementing the new denoiser, especially considering the time you have and from scratch.
+1
Before reading this I’ve thought that you implement some tech from outside…but now I know the full story… well…again…another evidence of your technical skills and attention to make a solid solution for us.
Thanks a lot Chaos Team
I am interested in how this approach could realistically evolve - before we invest in any new hardware. For me, starting multiple denoise jobs on multiple nodes, using skip existing, sounds workable via the command line. But needing to copy all the frames to each node first will likely add another big time overhead. Is it likely to be changed so that the denoiser can work on frames on mapped drives? As I write this, I think I must be missing something - how can skip existing work when the denoised output is saved locally?
Results from my first sequence denoise:
330 frames multichannel exr @ 1280x720, c. 86Mb each
Denoise took 2.5 hours on a Nvidia GTX 780, which we are pretty happy with: an average of 27s per frame that produced a cleaner result than the previous frames we had rendered, that took an hour longer each.
As an aside, has anyone done a GPU denoise on a large (6K+) render yet? If so, what hardware? Our large renders typically have many elements and whilst I understand that only the RGB is denoised, I figure the GPU will need to load a lot, if not all, the elements into memory first? Please put me right, if I am being dim.
It works fine with mapped drives; it didn’t work with UNC paths, but this is already fixed for the 3.40.02 patch. I also added a -frames option that allows you specify which frames from a sequence you want to denoise so you can distribute the work more efficiently among many machines.
Denoise took 2.5 hours on a Nvidia GTX 780, which we are pretty happy with: an average of 27s per frame that produced a cleaner result than the previous frames we had rendered, that took an hour longer each.Heh that’s very good to read
As an aside, has anyone done a GPU denoise on a large (6K+) render yet? If so, what hardware? Our large renders typically have many elements and whilst I understand that only the RGB is denoised, I figure the GPU will need to load a lot, if not all, the elements into memory first? Please put me right, if I am being dim.We detect how much GPU RAM is available and we split the image into strips that are fed to the GPU one by one, so we can handle quite large images - we don’t have to upload everything to the GPU at once.
Best regards,
Vlado
Thanks for clarifying the mapped drive support Vlado. Is that only in 3.40.02? Is there a release looming for this version?
Great news on the strips fed to the GPU also!
No, mapped drives have always worked.
Is there a release looming for this version?Alex queued the builds an hour or so ago, so we should be uploading them on the website later today.
Best regards,
Vlado
While the updated Maxscript UI has been uploaded yesterday, so feel free to grab it (from the usual place).
It’ll somewhat shorten also command-line work, as a control+click on the denoise button will simply copy the command line to clipboard.
If your using denoise a lot it is worth getting an AMD card for, at work i have a nvidia and a typical render takes 30 seconds to denoise during which you can’t do anything as the machine feels like it has frozen.
At home I have AMD and it takes 2 seconds to do it and you don’t even notice.
Fast forward 4 years >>>
Can anyone point me at an updated list of denoiser hardware benchmarks?