New GI in mental ray directions
If you haven’t had a chance to take a look at the new GI that is accelerated on the GPU, you can get a preview of the prototype here and here. ARC is also interested in your feedback, look here.
Also, if you haven’t had a chance to wander the ARC forums, here’s a little tidbit from Steffen, the project manager for mental ray about the future of Global Illumination in mental ray.
Going forward, what is now called “GI GPU” will be incorporated into a general new GI workflow in mental ray that can be GPU accelerated. It is supposed to be the modern replacement for the collection of various older GI techniques, like FG Map, Irradiance Particles and Importons, and GI Photons. If you have a capable GPU then the compute time can be dramatically reduced.
This means a single GI solution for all your needs in mental ray without the need to combine or tune other options to get the best result. Additionally it will take advantage of an NVIDIA GPU on your machine (if you have one) to drastically decrease your render times locally and/or a similarly equipped render farm.
Posted on June 13, 2014, in Uncategorized. Bookmark the permalink. 11 Comments.
You can preview GI on your GPU then bake, and render on your CPU ?
You can’t currently bake it. Part of the point of the new GI is to produce a fast and accurate brute force-based result that can be interactive. If you make it bake then caching data phase destroys the interactivity. Renderers are moving away from baking schemes in favor of faster/progressive feedback.
Suppose you want the CPU to do the render, that is why I brought up baking the GI, well not so much baking as taking the settings from the GPU GI to the settings of the CPU GI !
The new GI is either CPU or GPU so it is the same settings for both. The CPU completes the render in mental ray. This is not GPU rendering a la iray. The computation can be handled on the GPU which can generate the data much more quickly but isn’t required.
Oh, multiple GPU or a single powerful GPU is over shadowing the CPU !
Sort of. It’s doing the work for the CPU while it shades the scene. The CPU can be used for this as well if there’s no GPU available or your scene requirements (like geometry) won’t fit on the GPU.
This is a new technique. Not like the ones it is replacing. Hopefully more will be revealed at SIGGRAPH
When you say, if there is no GPU available, don’t most computers need a GPU ? I know there are situations where GPU are not available, I suppose my mind is drawing blanks 🙂
Render farms use blades that don’t typically have anything other than integrated graphics.
That’s what I thought. 🙂
Pingback: GI GPU Prototype Testing | elemental ray
Pingback: New GI Prototype Quick Start | elemental ray