Announcement

Collapse
No announcement yet.

Need suggestions on what to do with this...

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

  • Need suggestions on what to do with this...

    So this is 1 of 30 machines.



    1 machine has 8 CPU's with 6 cores on each CPU with 128 gigs of Ram.

    I get to render on them tomorrow. I will probably run Maya batch with V-Ray slaves on two host machines with Maya, and use V-Ray slave to handle the DR rendering.

    My problem was when I went to test out 3 of these machines with DR V-RAY RT. It would just sit for 15 seconds or more and then just give me a finished image. I know the network plays a huge part, but if there is anything I can do to optimize settings or try out, let me know.. I just get 1 shot of this.. I will record some video's, I just want to make sure any advice on optimization are observed for using a beast setup like this..

    This is from the Silverdraft truck. It's a mobile production studio on wheels..

    http://silverdraft.com/ Pretty insane huh?

    -Scott

  • #2
    I ran into some issues when testing. Figure I would report some info.

    There was no Rendering que software at the time for me to try. So my only option for rendering sequences was using Maya batch with V-Ray in DR mode.

    Problem I ran into was memory. The host workstation machine was 24gigs of ram with 8 cores. When it would use DR to render, those 48 cores per machine would only use 24 gigs of ram at most since it was mirroring what the host had even though they had 128 gigs. So I think I would have needed 128 gigs locally on the workstation, or run Maya in DR mode from a render node, but the Maya was node locked to the workstation.. I would not be able to run maya from those machines.


    Also, V-Ray in Maya batch mode rendering would say render complete after a few frames. I would check the log and it said it couldn't write the image file, but it would write the image file, it would just stop. Not sure why.. I don't have this issue at home rendering..



    The build was Stable build 16365.


    Hope this info helps.

    Comment


    • #3
      Originally posted by Metzger View Post
      My problem was when I went to test out 3 of these machines with DR V-RAY RT. It would just sit for 15 seconds or more and then just give me a finished image. I know the network plays a huge part, but if there is anything I can do to optimize settings or try out, let me know..
      You can play with the bundle size and samples per pixel in the RT settings to see if lower or higher settings make a difference. My guess is that increasing the number of samples per pixel may help - then each thread will have to work harder to finish its piece of the work and will reduce the amount of communication.

      Best regards,
      Vlado
      I only act like I know everything, Rogers.

      Comment


      • #4
        Originally posted by Metzger View Post
        When it would use DR to render, those 48 cores per machine would only use 24 gigs of ram at most since it was mirroring what the host had even though they had 128 gigs.
        If you mean the dynamic memory limit setting, it's on the "to do" list to be able to control it with an environment variable and also to compute the value automatically based on the amount of physical RAM on the machine.

        Also, V-Ray in Maya batch mode rendering would say render complete after a few frames. I would check the log and it said it couldn't write the image file, but it would write the image file, it would just stop. Not sure why.. I don't have this issue at home rendering..
        What image format were you writing to? What was the integer error code in the message?

        Best regards,
        Vlado
        I only act like I know everything, Rogers.

        Comment


        • #5
          Yeah, I think it would really haul if it was able to load up proper ram for all of those cores when the host machine has less ram in DR mode. I am going to try rendering again sometime to those machines via Deadline single frame submissions.

          The Image format was multi channel exr. The mayarenderlog error was: can't write image to disk.

          But it would actually write to disk, and it would say render complete.

          Thanks!

          Comment


          • #6
            I'm going to contribute nothing to thsi thread (sorry!),

            But that is insane, have fun!

            "1 machine has 8 CPU's with 6 cores on each CPU with 128 gigs of Ram."
            Maya 2020/2022
            Win 10x64
            Vray 5

            Comment


            • #7
              Same here, god damn I wish I could have some of that power when I need it

              Comment


              • #8
                This is the future: Render Gipsy! Always on the road...
                Portfolio: http://www.cgifocus.co.uk

                Comment


                • #9
                  Originally posted by Yannick View Post
                  This is the future: Render Gipsy! Always on the road...
                  LOL!!! Hehehehe

                  Comment


                  • #10
                    Hey Vlado,

                    I am back at the SilverDraft the truck (Render Gipsy). I have some questions.

                    Currently on a single 6 core machine, my render takes up around 16-20 gigs. When I submit a render to a 8CPU 48 Core machine with 128 gigs of Ram, I would think it would take more ram because of the extra cores. This isn't the case though and my render doesn't seem to improve speed wise. The ram never goes above 16-20 gigs. Do you know a safe build of V-Ray to test out? I am currently using a 2 month old Stable build. Any suggestions to try would be great!

                    Thanks!

                    -Scott

                    Comment

                    Working...
                    X