Originally Posted by sentry66
what exactly is the difference between this and iray?
Seeing as I did the presentations for 3ds Max I can feel comfortable answering this.
There are a few things that I think set the Visualizer apart from iRay/VrayRT.
+ It's multi-GPU speeds with the reference raytracing card but without memory limitations. So millions of polygons and GBs of textures.
+ Latency is very low. So you can modify topology and transform objects like you do in a raster viewport. Essentially no waiting as the acceleration structure rebuilds. You're looking at rebuilds per second instead of seconds per rebuild.
+ Its shader language is powerful and easy to use. The advantage of course being that they can add new shaders very quickly. In the last month or so they've gone from supporting a couple hypershade nodes to almost all of them. If it becomes popular it would be trivial to emulate loads of 3rd party plugins very quickly.
+ It supports arbitrary AOVs. So you can output normals, depth etc as sub passes (currently admittedly disabled to speed up viewport interactivity but the code is there).
I would also point out that while it's progressively rendering fPrime is a better analogy than iRay. While the Visualizer can be a path tracer (depending on your settings) it can also be an adaptive renderer like VRay, mental ray, brazil etc. So you aren't bound to physical effects like in iRay which for the most part locks you into physical accuracy.
But it's obviously not fully baked yet either and not ready for everything you would need from your primary production renderer.
Cons in its current version:
- You don't get true 3D motion blur with 4D sampling.
- There are some limitations with some use cases for intersecting SSS objects.
- No instancing.
- No Per-Pixel displacement.
- No implicit hair shapes.
- It's a non-blocking raytracer... which is really difficult to explain and would require a full post in of itself. But it's the cause for some of the challenges in intersecting SSS objects and makes some traditionally easy shaders a little more complicated.
Some of those can be worked around in many cases (multipass motion blur akin to iRay or Octane's, geometric displacement at render time, meshed hair primitives) but you can also of course mix and match. The idea is that it matches your final render closely enough that you aren't making any special accommodations in your scene for it. So if you have a character that needs per-pixel displacement and true 3D motion blur... render the background pass in Visualizer and then let the farm crunch the character. This is generally how I already work. I even still use scanline for mask passes etc from RPManager.
It definitely won't be for some people and it definitely won't be for every project but I'm genuinely pretty excited about it and have found it really useful, especially in product rendering work.