Why is Arnold too slow comparing to other CPU renderers ?

[David_Kup]

I even don't mention to GPU renderer. All other CPU renderers I know are faster than Arnold. Even faster than 3, 4 times. Like Vray next, renderman 22.3, Corona, Mantra. All of them are much faster than Arnold with same quality. And people are moving to others day by day. Why ? Why doesn't Developers focus to speed ? I hope Deveplovers and Direction know this situation and get reasonable Roadmap. Thanks so much !

[Christoph_Schaedl]

In complex animated scenes the speed comparison isnt that much of a difference. And speed isnt the only reason to choose a render.

But i agree it could be faster.

[schwungsau]

I am using most of the renderer in production in past years. I found out nothing is slower then Mantra, it outdated, that why sidefx will release Karma to replace it. renderman22.6 has a similar speed like Arnold. Corona and Vray are faster because of its "biased" rendering. the whole point of "biased" rendering to be faster then "correct inbiased" light tracing, but the setup is more complicated, you need to know the renderers optimization techniques to maintain the quality.

also, you have to keep in mind, some techniques are better or worse for different kinds of scenes. Indoor scenes or Caustics are very bad with Eye Pathtracer like Arnold or Renderman. Vray/Corona is fast because it uses light caches, photomapping. etc to help it out. the best solution would be a bidirectional path tracer or MTL like renderman-unified integrator or indigo renderer.

[T0M0X]

Arnold is slow ? Then you never worked with Maxwell...

[hyperviktor]

Let's not compare Corona to Arnold, if rendering speed is your only comparison you shouldn't be doing CG or VFX.

In any bigger production, rendering speed is pretty much irrelevant, the renderfarm will chew through anything. We're basically talking about 50 hour frames sometimes, how does that sound ?

And you really need to do proper comparison, Arnold is capable if handling anything - you throw at it hundreds or thousands of lights, millions or billions of polygons and it will still be fine and happily rendering.

As long as you don't have complex scenes where your render times are hours per frame, you can't really tell how these renderers perform. And even then you don't want to compare the same scene, you compare render results.