7 Replies Latest reply on Aug 15, 2016 9:27 AM by scotth97401340

    Awful performance on Raytraced GPU renderer vs. CPU for depth of field

    scotth97401340 Level 1

      I'm experiencing about a 100x performance decrease using the Raytraced GPU renderer compared to Classic 3D CPU rendering, and I'm hoping someone might be able to help me diagnose why.

       

      My GPU is an Nvidia Quadro M4000. The CPU is an Intel Xeon 10-core at 2.3Ghz. The machine is running Windows 10 on 32GB of RAM.

       

      My composition is 4K resolution. The scene I'm rendering is very basic: a set of flat text layers arranged in space as the camera moves past them with depth of field enabled. The effect is that the text goes from out of focus to in focus, then back out of focus again. Pretty simple.

       

      With Classic 3D rendering, I can preview at 5-10 fps, depending on the particular section of the scene. But with Raytracing, each frame takes 10s or more (and also looks like crap, but that's a different issue). My ray tracing quality is set to just 3, and I've got Anti-aliasing set to Cubic. The composition is set to use 8 samples per frame for motion blur.

       

      So, what's up? Is the Quadro M4000 just a really bad fit for this scenario? Would I be better off with a Geforce 1080 or a Titan X, or does it sound like there could be a configuration or driver issue that might be the culprit for such poor performance?

       

      Thanks in advance for your help!

      -Scott