We are a studio that has primarily used the CPU Production Renderer for all of our work. Over the last week we have been testing the CUDA Renderer with 2x 2080 TI’s with a view to looking at if rendering on GPU’s would offer any benefits to us.
Our findings are somewhat overly simplified but generally the following;
Rendering on the GPU’s is almost twice as fast as on the CPU (comparing 2x 2080 TI’s to 2x Xeon E5-2699 v3’s) but comes with the compromise of less features, some of which can be deemed an integral part of existing pipeline / workflows.
It would be useful to know what the roadmap for the GPU Renderer looks like and if we are going to get to a point where we have most, if not all, of the same features on both renderers?
For the moment we are focused on several big features: RTX support, out-of-core geometry, OSL support and deep output. All of them require substantial amount of effort, but I’m glad that there is very good progress. We will show more at Siggraph in about a month or so.
Beyond that, if there are any features that you find missing for your workflow, please let us know so that we can priotitize them.
So there is no road map at all?
is there a list of things you want to implement besides what you are working on right now?
it would be good to know, so people don’t separately request and nag all the time (including me)
Of course we have a very long list of things that we want to implement, but I specifically want you to hear from you - it helps a lot with prioritizing stuff.
I’m usually very careful about posting future plans and only post about things that I’m pretty sure about; on the one hand I don’t want to overpromise, but on the other hand people get discouraged if they don’t see their favorite feature at the top of the list. That’s why I would prefer to hear your feedback and internally adjust the plans accordingly.
For us, a few of the key features of V-Ray Production, that we rely upon for most of our projects are as follows;
Material Select Render Elements
Trace Sets Support
Full Cryptomatte support
VRayScannedMtl support
Dome Camera support
Object Select Render Element support
Total Light Render Element support
That’s all I can think of from the top of my head. Of course, ultimately our preference would be a like for like render engine as much as is possible. Having the real ability to use either CPU or GPU would be great
Refraction affect channel functionality.
The ability to exclude geometry from the ExtraTex pass.
As mentioned above, bump maps between Next engines.
A standard material library like corona and other engines would be helpful, I’m not sure why this hasn’t been done before but would be very helpful.
Lightmixer! Would be great!
- Full Phoenix particle shader support to enable foam/splash (Instead of the current VRay particle workaround)
- Bump map standardization between CPU and GPU
- Custom mapped density in VRayEnvironmentFog
- VRayDistanceTex
- Full Phoenix particle shader support to enable foam/splash
I’d like to see that I tested yesterday with only one RTX2080, and the ocean scene render was extremely fast! I would love to be able to render out the foam and splash also on GPU.
I would add the VrayMatteShadow render element to this list. I still use kick out 2 AO passes (one at 2 inches/5cm radius and another at 1 foot/30cm) and I use the Matte Shadow element as a mask so that I only get AO in the shadows. The VrayShadows element is not a viable replacement because I am doing a great deal of solid stainless steel renders and the lighting/shadow data is very limited because metal has no diffuse values.
