The devil is really in the details. Pixar and Disney throw scenes at Renderman for overnight rendering jobs where each node loads up to terabytes of data per fr

The devil is really in the details. Pixar and Disney throw scenes at Renderman for overnight rendering jobs where each node loads up to terabytes of data per frame (geometry, textures, volumes...). So the renderer contains tons of tricky optimizations under the hood to support that crazy amount of data. Other renderers like Cycles use pretty much comparable physical models for everything and create almost identical output for simple scenes. But most of them would crash and burn spectacularly of they had to deal with scenes at the size these studios produce. [Even though you could theoretically generate a lot of detail, especially textures, algorithmically on the fly, nobody goes for that: the only known good way to band limit textures for level of detail reduction without aliasing artifacts is to sample and filter them - generating the whole texture anyway.]