Announcement

Collapse
No announcement yet.

Ship on the ocean sim issues

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

  • #16
    Roger that, got sidetracked by other tasks, will try to get back to this one and finish it in a few days.

    Cheers!
    Svetlin Nikolov, Ex Phoenix team lead

    Comment


    • #17
      Svetlin,
      Thanks for your attention on this. I am looking forward to a run of about 3000 frames.
      Ron

      Comment


      • #18
        Hey, I am finally back on the subsiding liquid issue. Will ping you when there is progress.

        Cheers!
        Svetlin Nikolov, Ex Phoenix team lead

        Comment


        • #19
          Hey, the fix for the subsiding liquid will be in tomorrow's nightly builds, sorry for the long delay!
          Svetlin Nikolov, Ex Phoenix team lead

          Comment


          • #20
            Thanks for this great news! I will test and let you know the results.
            Ron

            Comment


            • #21
              Update:
              PhoenixFD 3.12.01 Nightly 28817, VRay Next 4.10.01, 3dsMax 2019.3 Update. This is frame 350, sim time for this frame 17 minutes on i7-3770@3.40GHz with 32GB. 84 Million cells, 50M Liquid, 13M Foam, 85K Splashes. This frame cache file size 520MB. Render for this frame 1920x1080 time 23 minutes on BB DR farm 5 each i7-3770. Majority of bucket time is spent on wake foam and slash with blur x2. Lighting = sun + VRay dome light with hdri.
              Strong rotating emitters same size as the ship screws are creating the underwater foam, pattern is set very low 0.04, and droplet surfing is 1.0. Screw emitter force is 20 feet per second faster than ship speed of 25 knots.
              Render time is okay for stills, not really practical for animation at this quality level and resources.
              No more problems with diminishing ocean in the front of the container, that I can see at this point, and not seeing much of the surface scratching with displacement detail 10 and strong wind. With sim time 4 frames per hour I can let this run a few more days then try something else, perhaps excluding the ocean from the dome light.
              Would changing the grid scene scale to a larger value help sim time? Current scene scale is 1.0.
              Suggestions to improve speed are welcome. Sim file attached.
              Attached Files

              Comment


              • #22
                Hey, asking just in case - if your foam or splash are rendering as bubbles/cellular/splash and you have the Liquid Simulator linked in any of the Particle Shaders, you certainly wanna try a nightly 4-5 days newer - it might benefit the render speed of the particles a lot.

                In any case, you should boost your Light Cache Speedup of all Particle Shaders to as high a value as possible until you start getting render artifacts - this should also help the render speed a lot.

                There is a workflow which would help render a lot faster, because the Particle Shader rendering is often bound by the huge lot of refractions going on at the liquid surface when the particles are floating around it. You could render in 3 separate passes the ocean mesh, the particle subgroups above the liquid, and finally the particle subgroups below the liquid surface, taking care to set the particle shaders' liquid tint and turbidity options to simulate the fog of the water material. The sum of the time for these 3 passes should be significantly lower than rendering the scene in 1 piece, for this kind of setup. If you want to go that way, give it a try with a short sequence - there will be a certain visual difference because refractions will be omitted and reflections of the underwater foam from the ship won't be there, but these should be negligible/

                I still have to finish the issue with sim slowdown before restore - this should help shave of some 20-30 or even 40% of the simulation time for this scene too...

                Cheers!
                Svetlin Nikolov, Ex Phoenix team lead

                Comment


                • #23
                  Thanks Svetlin!
                  Rendering the frame above was with both foam and splash shaders in point mode and speed up set 0.90. Not using mist or wet map.
                  This is simulated and rendered with nightly 28817 from December 31. I saw the bug fix for slowdown of particle shaders was done a few days before on December 28. Not seeing anything on the change list for the January 2 nightly.
                  Thanks for the workflow tip! The attached render with above-surface foam took 6 minutes with global illumination turned off and speedup set to 0.99. The ocean mesh and sub-surface foam were very fast. Now I get to learn how to comp these layers together. Fun!
                  Ron


                  Comment


                  • #24
                    Indeed, I missed adding it to the changelog initially and now there seem to still be issues in that thread, so it should better not go into the changelog until we're sure what's going on.

                    Btw, don't hesitate boosting the speedup even to 0.999 - with my scenes sometimes even that number is good, and sometimes I have to lower it back to 0.996-7, but the difference between 0.99 and 0.999 might still be significant (because the remaining number to 1 is actually the important value and the it's 10-fold between 0.01 and 0.001).

                    Cheers!
                    Svetlin Nikolov, Ex Phoenix team lead

                    Comment


                    • #25
                      Update:
                      Speedup set to 0.999, reduced diffuse multiplier on foam and splash shaders to 0.8 and changed color of foam and splash points from white to 200, reduced shadow subdivs of the sun, GI is turned off. Bucket image sampler with 3 min / 100 max / noise threshold 0.07, following the "universal" workflow tool-tip on V-Ray Next.
                      Result is 6 minute single pass render on my farm, which I am very happy with. There may be other improvements possible but this is the best result I have seen to date.
                      Only issue remaining is the sim time slowdown. Hoping you have success on that!
                      Thanks again Svetlin for all your help!

                      Ron

                      Comment


                      • #26
                        Yay, that's much better! Hopefully during the next months there will be progress on the slowdown issue as well.
                        Svetlin Nikolov, Ex Phoenix team lead

                        Comment

                        Working...
                        X