Announcement

Collapse
No announcement yet.

smarter bucket size

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

  • smarter bucket size

    why its not possible to modify the bucket size during render (split up last buckets)

    or start render with buckets size 64x64
    at 30% of render buckets are 48x48
    50% is 32x32
    to 90% 8x8

    render first 50% "checker order" and during last 50% we assign most heavy "neighbour" buckets first

    or use GI sample density
    zdeept
    poligon count
    ..?..
    and assigne that reagions first

    that way we could busy all core during render and not waiting for the last bucket
    __________________________________
    - moste powerfull Render farm in world -
    RebusFarm --> 1450 nodes ! --> 2.900 CPU !! --> 20.000 cores !!!
    just 2,9 to 1.2 cent per GHZ hour --> www.rebusfarm.net

  • #2
    Yep, this is somewhere on the to-do list

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

    Comment


    • #3
      But, will there be any speed improvements doing this ? or will the complete rendering time be the same ?

      Comment


      • #4
        It will be slightly faster I guess.

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

        Comment


        • #5
          and what is on the todo list ? split last bucket or shrink buckets or something other secreat
          __________________________________
          - moste powerfull Render farm in world -
          RebusFarm --> 1450 nodes ! --> 2.900 CPU !! --> 20.000 cores !!!
          just 2,9 to 1.2 cent per GHZ hour --> www.rebusfarm.net

          Comment


          • #6
            nice...progressive bucket size render

            Comment


            • #7
              I suppose it would use the bucket size you specify until there are fewer remaining buckets of that size than cores/threads available, then it would start splitting so that all cores are involved, down to a certain size, eg: You have 8 cores and spec 128x128 buckets, so Vray renders until only 7 cores are used, then divides reamining pixels by 8, etc. This might be limited because calculating the AA at the edges of very small buckets may take longer than it would take to render bigger buckets, even though all cores are not used.

              Comment

              Working...
              X