The standard mode of mental ray is called technically ‘distribution ray tracing’. It means basically in easy words that the renderer samples lights, materials, depth of field, motion blur each of that independently from each other (ie. at each intersection, rays are distributed over an interval.. soft shadows by distributing rays over an area light, blurry reflections over a solid angle and so on). That means a lot of time to wait for complex scenes. The more complexity you add the more your render times go up. That means to deal with today complexity a distributed raytracer can barely keep up with other modern renderer. That’s why mental ray can also do Progressive ray tracing with its Unified Sampler. Unified sampling in Progressive mode is very very powerful and keep mental ray up to the industry standards. In facts many of the complex scenes you see here have been rendered on Progressive.
You can look at Standard mode as the past, and at Progressive as the future.
Our sampler is able to reconstruct the various sampling calls from a lens while sampling DOF, acting like a global sampler but as soon as you have to render reflections of reflections over shadows and shadows and depth of field etc. you’ll get to the edge. It’s really a technology limit. We wanted to dev our sampler to really put an end on this. You can’t go probably much further with a distributed renderer and an ad-hoc adaptive sampler like we adopted for Rombo. There’re some tecniques we’re looking into based on lightfields technology that can tell us where the DOF blur is so we can choose different approaches based on that. That may improve speed and consistency. However if you need more complexity then you’ll have to go Progressive.
That’s why we fully support it. Because it’s not only about importance sampling here that looks like the mood of the moment. There are coming out a lot of things that keep pushing this kind of renderers. And we’ll keep adding them up as soon as they are reliable. For example our Rombo.Material is doing something very new and powerful to converge in less time for more realistic results. And we keep doing that because at the end of day we wanna our renderings finished and looking great.