Announcement

Collapse
No announcement yet.

Build a render farm, or wait for CUDA?

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

  • Build a render farm, or wait for CUDA?

    Hello! As I mentioned on my last post, I am seeking to improve the speed of my workflow and also my rendering production. I want to implement RT in an Active Shade for my large architectural scenes. Should I invest in i7 nodes, or wait for possible big announcements in CUDA/RT? From my research, it sounds like something new is coming from RT and CHAOS GROUP? Thanks!

  • #2
    Cuda is here.

    I go with i7, GPU rendering dont give u render passes/elements (or I mist something?) so GPU rendering for production is quite useless...

    GL
    CGI - Freelancer - Available for work

    www.dariuszmakowski.com - come and look

    Comment


    • #3
      Thank you very much... I did not know if we were on the cusp of seeing VRAY RT ready for production. Can you advise me for a solution for Active Shade? Is there a formula for poly count to derive how many gb you would need on the GPU? (3,4,6gb?) Or, should I use my rendering nodes for Active Shade?

      Comment


      • #4
        Heya

        Its hard to advise out of thin air, I would need to know roughly what u work on and how to advise... Basically speaking u are looking at 100mb of GPU memory per 1 million polygons (or at least that what was 1 year ago).

        Every 2 GPU based graphic card has only 50% of its Memory usable. They dont share memory which means u have to fit all in half so that each GPU can work on scene.

        Also GPU rendering is missing a lot of features and as far as I know textures are still down sample (its soon to be changed...) So its tricky one to use GPU for production...

        The only reason I would use GPU is to do a session with client, set the light mood etc etc what he wants to see and then work on CPU...
        CGI - Freelancer - Available for work

        www.dariuszmakowski.com - come and look

        Comment


        • #5
          Originally posted by DADAL View Post
          Heya

          Its hard to advise out of thin air, I would need to know roughly what u work on and how to advise... Basically speaking u are looking at 100mb of GPU memory per 1 million polygons (or at least that what was 1 year ago).

          Every 2 GPU based graphic card has only 50% of its Memory usable. They dont share memory which means u have to fit all in half so that each GPU can work on scene.

          Also GPU rendering is missing a lot of features and as far as I know textures are still down sample (its soon to be changed...) So its tricky one to use GPU for production...

          The only reason I would use GPU is to do a session with client, set the light mood etc etc what he wants to see and then work on CPU...
          Thanks, Dadal - I appreciate the assistance! I am going to focus on building a CPU render farm - that will be able to host GPU's in the future.

          Comment

          Working...
          X