Announcement

Collapse
No announcement yet.

Understanding DMC Sampler

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

  • Just taking a look now Stan, some initial observations:

    Are you going to do a lot of work in post? If not, why not just bake the gamma in to the image (uncheck don't affect colors in color mapping, and untick sRGB in the VFB) that usually renders quite a bit quicker in my experience. [EDIT: ignore that statement about it rendering quicker, the thing that makes the difference is the bit below]

    Also, even if you do keep 'don't affect colors' ticked, I don't see any reason for not using reinhard color mapping and lowering the burn value to something like 0.25. On a (not very) quick test that alone took the render time down from 23mins to 14 mins.

    For an explanation from The Man himself look here: http://www.chaosgroup.com/forums/vbu...579#post543579

    Will continue going through it when I have time.

    Peter
    Last edited by peterguthrie; 27-06-2013, 07:17 AM. Reason: incorrect info
    www.peterguthrie.net
    www.peterguthrie.net/blog/
    www.pg-skies.net/

    Comment


    • Humm, thanks for that feedback Peter,
      Well, we are used to go out with linear at work, and so it's more a theoretical limitation.
      But yeah, I'm not against going to reinhard if it's the only way to get it out.

      But, isn't it the same to go to reinhard and tick the "don't affect color"?
      That tells Vray to save as linear but to sample as Reinhard. That means it should render the same speed as the normal Reinhard, but you can still add and substract and having the power of linear in comp. The highlight will be not really well aliased, but who cares, specially if we do the Reinhard curve in comp, then we end up with the same output and same rendertime but with the LWF power, isn't it?

      I will have a look to Vlado's post when I have time, really interesting, I'm in a middle of a rush now (why am I on the forum anyway then?)

      Stan
      3LP Team

      Comment


      • Originally posted by 3LP View Post
        But, isn't it the same to go to reinhard and tick the "don't affect color"?
        I did two renders, both had "don't affect colors" ticked, but one used reinhard burn 0.25 and the other didnt. The reinhard one was 14 mins and the other was 23mins, so definitely not the same speed.
        www.peterguthrie.net
        www.peterguthrie.net/blog/
        www.pg-skies.net/

        Comment


        • Now that is very interesting... I always thought that the reinahard/linear/expo etc etc were only post process effects and did not affect render times too much. I gotta revise my knowledge now.

          Thanks for info!
          CGI - Freelancer - Available for work

          www.dariuszmakowski.com - come and look

          Comment


          • Originally posted by cecofuli View Post
            Ohh... welcome back GoncaloP !!!
            Thank you!

            Comment


            • Yeah sure,
              Sorry if I wan't clear, I meant that if you rendering "don't affect color" ticked and reinhard, it would be the same speed as reinhard straight away.
              Sorry we are on the same lenght here

              All right, +1 for that, what else?

              Thanks

              Stan
              3LP Team

              Comment


              • Originally posted by 3LP View Post
                Yeah sure,
                Sorry if I wan't clear, I meant that if you rendering "don't affect color" ticked and reinhard, it would be the same speed as reinhard straight away.
                Sorry we are on the same lenght here
                yeah probably should I guess
                www.peterguthrie.net
                www.peterguthrie.net/blog/
                www.pg-skies.net/

                Comment


                • Stan,

                  These are all still using universal settings with a high noise threshold because I don't have time to do longer render.

                  This one more or less without any changes to your settings: (there are chairs missing because of a missing plugin, and I havent turned on the interior lights yet)

                  Click image for larger version

Name:	a.jpg
Views:	1
Size:	316.0 KB
ID:	847818

                  23:36

                  And this one with reinhard mapping and burn at 0.25:

                  Click image for larger version

Name:	b.jpg
Views:	1
Size:	344.3 KB
ID:	847819

                  14:27

                  I noticed your light cache settings were a bit odd, so I put things back to default but also ticked use LC for glossy rays:

                  Click image for larger version

Name:	c.jpg
Views:	1
Size:	316.6 KB
ID:	847820

                  3:59

                  So that is all before even messing around with DMC settings and subdivs etc. which I will try next.
                  www.peterguthrie.net
                  www.peterguthrie.net/blog/
                  www.pg-skies.net/

                  Comment


                  • On the interior lighting, if you can at all bear it, excluding the lamp itself from shadow casting makes a big difference. I usually do this in my scenes as everything just goes way too slow otherwise. It means the light distribution isn't correct, but given it renders 2x as fast I would personally live with it. (I lowered the multiplier as well as it didnt need to be so bright when not being obstructed by the shade)

                    lights turned on:

                    Click image for larger version

Name:	lighta.jpg
Views:	1
Size:	322.5 KB
ID:	847821

                    17:05

                    lights set to exclude lamp object (cast shadows only):

                    Click image for larger version

Name:	lightb.jpg
Views:	1
Size:	317.1 KB
ID:	847822

                    9:54
                    www.peterguthrie.net
                    www.peterguthrie.net/blog/
                    www.pg-skies.net/

                    Comment


                    • Nice one Peter!
                      All this is very interesting!

                      So you are still using the HDRI, I wan't expecting anymore seeing everyone just saying that it was to hard

                      Thanks,
                      Stan
                      3LP Team

                      Comment


                      • Originally posted by 3LP View Post
                        Nice one Peter!
                        All this is very interesting!

                        So you are still using the HDRI, I wan't expecting anymore seeing everyone just saying that it was to hard

                        Thanks,
                        Stan
                        yeah still with the hdri. What size was your 12 hour render?
                        www.peterguthrie.net
                        www.peterguthrie.net/blog/
                        www.pg-skies.net/

                        Comment


                        • Reading from a 3LP post, it was a 700x900 that tooked 12h with still noise.
                          I'm quite curiouse to see what render times you wil get Peter.

                          Keep it up
                          Gonçalo

                          Comment


                          • Originally posted by peterguthrie View Post
                            Tried a new test over the weekend:

                            Through trial and error I found 24 subdivs to be the fastest
                            That's strange. I just applied your (new) settings to a test scene of mine. But in my case the increase of dof subdivs made the render faster and the samplerate channel darker. Even went till 100subdivs :O


                            Oh by the way. These are killer settings, rock on!
                            Nils Poetoehena
                            3D Visualiser
                            www.demanufacture.org
                            www.gielissen.com

                            Comment




                            • This is what I was able to get within 1 hour range using 2 single computers Intel Xeon W3520 @ 2.67GHz, 12 GB DDR3 @ 534MHz


                              Settings were:

                              Adaptative DMC image samples Min/Max 10/128, Clr thresh:0,05

                              Color mapping: Reinhard - 1,0; 0,0; 2,2 (Sub-pixel mapping + Affect background Ticked)

                              Primary bounces - Brute force 512 Subdivs
                              Secondary bounces - Light cache with Vray defaults and Use light cache for glossy rays and Retrace threshold 1,0 ticked

                              DMC Sampler:
                              Adaptative amount 0,7
                              Noise threshold 0,005
                              Min samples 16

                              Material subdivs 128 for the overall image

                              Among some small details, The Light balls glass refraction seams to be the main killer here and for glossiness got the help of interpolations and no affect shadows option

                              Thanks to 3LP that sent me this scene!
                              I will send you the scene back later tonight and try write some more deep explanation as well.
                              Last edited by GoncaloP; 01-07-2013, 05:51 AM.

                              Comment


                              • Hi 3LP,

                                I don't have much experience with HRDI maps and I would like to have your help to understand the values you use in your scene.
                                Overall mult: 0,025 and Render mult: 1000 of the Peter's HRDI map, are those recomended settings? What should they produce? Can you explain me why you use those instead of default 1,0 for both?

                                Thank you

                                Comment

                                Working...
                                X