Announcement

Collapse
No announcement yet.

Out of core

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

  • Out of core

    Does RT support this or vray when doing gpu rendering?
    Architectural and Product Visualization at MITVIZ
    http://www.mitviz.com/
    http://mitviz.blogspot.com/
    http://www.flickr.com/photos/shawnmitford/

    i7 5960@4 GHZm, 64 gigs Ram, Geforce gtx 970, Geforce RTX 2080 ti x2

  • #2
    This has been discussed quite a few times here - accessing main system slows down the render to a state where there is no point of rendering on the GPU.
    What other renders are doing is making memory optimizations, so you can fit more on the GPU and calling this out of core. If you really push the limits, it will get very slow.
    V-Ray RT GPU has already some approaches on fitting as much stuff as it can. We have features for this with the next SP (most of those are already in the nightlies) - we will detect the real number of textures channels for example (even if it is png or jpeg), so if you have a monochrome texture for bump for example, it will use 4 times less memory. If the alpha channel is completely white, we will not reserve memory for alpha channel, etc. In the current version we always allocate memory for 4 channels.
    We are still working on the system that will also generate mip-map levels on the fly, so the texture will use as little memory as needed. This will also greatly reduce the memory usage, I hope it will be ready for the next SP.

    Best,
    Blago.
    V-Ray fan.
    Looking busy around GPUs ...
    RTX ON

    Comment


    • #3
      I have a couple of questions regarding textures and memory. Forgive me if they may seem obvious, but please bear with me and put them out in laymans terms

      1) raw pixels vs compressed images ?
      Does the compression on a file texture make any difference ? When put to memory on the GPU, does the source make a difference?
      ie. if you have an "identical" image saved as uncompressed, say, bmp - and another saved as jpg. Do they take up the same memory footprint once uploaded to memory?
      is it just all about bit depth and pixel count once in memory.


      2) Node combining file textures.
      If a source file texture is loaded into maya hypergraph twice, linking into two different shaders. Does it take up twice as much memory? or is it optimal to delete one and link the remaining into the two shaders?

      3) File referencing and textures.
      If a scene has other prop scenes referenced in, and they have nodes pointing to the same file texture. Are they intelligently merged?


      Again, apologies if these are "duh" questions, but as a non-programmer I had to ask.

      RH

      Comment


      • #4
        Well, I think these are actually very good questions.
        We need the textures uncompressed in order to be able read them during render, so it doesn't matter. If a texture is used multiple times in the scene, it is loaded only once, no matter how it is referenced

        Best,
        Blago.
        V-Ray fan.
        Looking busy around GPUs ...
        RTX ON

        Comment


        • #5
          Clear, concise, superb answer. Thank you Blago

          Comment

          Working...
          X