Announcement

Collapse
No announcement yet.

Cleanup noise in Elements: Reflection and Shadow - Also LOOONG render times

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #46
    Thanks for lending us your brain Dmitry
    When I eventually do the shops on the sides (which isn't the main focus but still need light from there to spill into the passage), would you also suggest to set the lights from those to also skylight portal like you did with the big vraylight behind the camera?

    That light doesnt work for me as a skylight portal as its way too dark then in that passage. I made it a "fake flash" to brighten up the passage way over at the back where the camera is looking from. When the shops are in, then this light's mulitplier can come down a bit, or even be removed completely since the lights from the shops should fill the passage.

    Whats your thoughts on this?

    nice clean renders from you by the way
    Last edited by Morne; 03-05-2012, 10:23 PM.
    Kind Regards,
    Morne

    Comment


    • #47
      if I make adaptive 0, then render times comes down further and so far I don't notice a difference in noise after a couple of buckets
      Kind Regards,
      Morne

      Comment


      • #48
        I dont understand what is going on here now. Now I want the opposite result. Don't care about the noise, but need FAST preview. Cant seem to get fast preview even if I purposely make BF settings LOW. changing adaptation to 1 or to 0 still dont result in fast preview
        Kind Regards,
        Morne

        Comment


        • #49
          just raisse noise threshold to 0.5.. should go fast then?

          Comment


          • #50
            I'm working on my main scene now
            even putting it on 1 doesnt do much
            Kind Regards,
            Morne

            Comment


            • #51
              I just use a script to globally set all the shader/gi to low for tests, and then revert it to high. Actually I got a pretty similar result with universal settings a bit more grainy and a bit slower.
              Dmitry Vinnik
              Silhouette Images Inc.
              ShowReel:
              https://www.youtube.com/watch?v=qxSJlvSwAhA
              https://www.linkedin.com/in/dmitry-v...-identity-name

              Comment


              • #52
                Here is my new attempt trying to merge Morbid approach to DMC sampler with my IM/LC workflow; it seems to work pretty good, maybe my results aren't so clean as his ones but it takes half the time with my gear. I suppose that using a fully textured model it could be possible to raise a bit the clr threshold (noise introduced should be invisible) in order to minimize rendertimes even more.
                Attached Files
                Alessandro

                Comment


                • #53
                  i'm curious to have a peek at Morbid's settings, would anyone be so kind as to post some screenshots of the settings? (I dont have max 2012)
                  www.peterguthrie.net
                  www.peterguthrie.net/blog/
                  www.pg-skies.net/

                  Comment


                  • #54
                    Originally posted by peterguthrie View Post
                    i'm curious to have a peek at Morbid's settings, would anyone be so kind as to post some screenshots of the settings? (I dont have max 2012)
                    here is the scene he posted, saved as 2011
                    Attached Files
                    Alessandro

                    Comment


                    • #55
                      Originally posted by zeronove View Post
                      here is the scene he posted, saved as 2011
                      thanks
                      www.peterguthrie.net
                      www.peterguthrie.net/blog/
                      www.pg-skies.net/

                      Comment


                      • #56
                        basic idea of my settings is:

                        set dmc aa sampling to 3/16 - 3, to get fine edges and lines properly aliased, 16 to get enough rays for the reflections at extreme angles.
                        use script to globally set all subdivs to 128 (for first test) that goes for anything that is seen in frame like glossy reflections/refractions, lights, gi (if its bf)
                        leave dmc sampler noise controls at default 0.85/0.01

                        At the test render you can still see some grain, not a whole lot. I typically look at each render element - raw gi, raw lighting, reflection/refraction to determine where the noise is most prominent.
                        After that I adjust the sampling for the noise areas, i.e. if everything is clean, but reflection has grain, up the samples in reflections for all materials. Reason being is that one material may get enough samples, but it may reflect another material that has grain, so therefore it will also have grain in it. I find globally raising all subdivs solves that quite easily.

                        Once Im happy with the look but overall the image still feels grainy (what I mentioned about measuring noise differences), I then lower the adaptive amount and adaptive threshold, until I get clean result, typically, if I have .85/0.01, .7/0.005 guarantees clean result (more or less )
                        Dmitry Vinnik
                        Silhouette Images Inc.
                        ShowReel:
                        https://www.youtube.com/watch?v=qxSJlvSwAhA
                        https://www.linkedin.com/in/dmitry-v...-identity-name

                        Comment


                        • #57
                          Thanks for sharing your approach Morbid Angel.

                          I am curious in what situation you will start cranking up the Max subdivs of Adaptive DMC from 16, instead of upping material subdivs? If you see grainy area with red sample rate in the RE, will you up the Max ADMC subdiv or the material subdiv or both to clean up the noise for smoother glossy reflection/inter-reflection between objects?
                          always curious...

                          Comment


                          • #58
                            I rarely touch the min/max, only extreme cases.
                            Dmitry Vinnik
                            Silhouette Images Inc.
                            ShowReel:
                            https://www.youtube.com/watch?v=qxSJlvSwAhA
                            https://www.linkedin.com/in/dmitry-v...-identity-name

                            Comment


                            • #59
                              Very interesting approach, thanks for explaining it as well, it's starting to make sense to me at last. I'm currently doing some tests and it looks like it compares very favourably to the universal method. (for the scene I am testing at least : quicker AND less noise) (I realise that doing vray tests on a friday night is maybe a bit sad )
                              www.peterguthrie.net
                              www.peterguthrie.net/blog/
                              www.pg-skies.net/

                              Comment


                              • #60
                                Ps. The really cool thing is being able to throw more samples at problem areas, which is essentially what I was looking for in my other thread
                                www.peterguthrie.net
                                www.peterguthrie.net/blog/
                                www.pg-skies.net/

                                Comment

                                Working...
                                X