After spending some time with iray I have finished something worth showing.
720p iray animated hybrid.
http://www.youtube.com/watch?v=K6pLfxGTsWw
Here is my review to go with it.
The Good
Love the physical nature of the renderer. Everything is always on. All the most advanced lighting features of mental ray are just, on.
Its incredibly fast. The above animation rendered in 1min 45sec per frame on average. It doesn’t seem to care about soft reflections, or color bleed, or refraction.
Simple scenes, it takes less tweaking to get things to look right. No occlusion pass, no facing ratio pass, etc.
Progressive style render is nice for tweaking.
The Bad
Batch is a bit crashy when doing lots of frames (no surprise here, its not really even released for maya)
No custom shaders (only MIA, at least its a good one)
BDRF doesn’t work the way I want it to. ( I cant push it very far )
No luck on rendering from service, It seems that you need to render it from a login that is driving a monitor to get access to the cuda display driver. Big deal for farms. I think this is nvidas problem, not mental images. I think it will come with time.
More Limited Memory, wasn’t a big deal here. My cards all have at least 1gb and that car scene never topped 500mb.
Conclusion
I can see why mental images didn’t want to fully release this yet. Its not quite ready. Its still an incredible piece of software and I really want to use it in production. But I wouldn’t use it on any long sequences. Some of the limitations that I thought would drive me crazy turned out to be the opposite. I like the simple way of working with iray. Sometimes it feels elegant instead of restricting. Things just render the physically correct way, you don’t have to do so many tricks to make things look good. You can still crank up your style in comp anyway. The speed is incredible. I am glad that I have a whole farm ready for video cards, but I am not going to buy any yet. Anyway, you could render that car shot in 6 hours on one machine. Thats pretty amazing.
Final Thoughts
If mental images can sort out the bugs and make it stable this will be my #1 rendering tool. It cant do everything, but its incredible at what it does. Pair that with normal mental ray and you can do just about anything.
If they can get motion blur, custom shaders (like SSS), good large memory management, and the reliability rock solid this could give Renderman some stiff competition for the best renderer in the world.
I am stoked.