I have an object like this. It's basically an icosphere with holes, which I got by selecting all faces, Select Menu > Checker Deselect and then S + 0.


The problem starts when I try to render it in Cycles. The inside is very noisy.

enter image description here

Increasing the Samples doesn't help.

Clamping doesn't help either. Nor does Multiple Importance in World settings. If I try increasing Diffuse Bounces, the inside gets overly lit, but the area close to the corner, which has the remains of a shadow, stays noisy.

In Blender Internal, I could go to Gather tab in World settings and set a Falloff.

enter image description here

And that yielded this result:

BI result

I'm not sure how realistic it is, but at least it's somewhat fast and noiseless.

But since BI is said to be obsolete and might be kicked out in the upcoming future, I want to know if it is possible to achieve similar results in Cycles. Also, I'm trying to bake lighting to a texture, so a way that supports baking as well would be great.

EDIT1: Posting results of renderings with different amounts of samples:

enter image description here

  • $\begingroup$ 50 samples is still rather low for most cases.. Try maybe 500. But you are right, I've noticed that AO is often still noisy after the full shaded scene has converged to an acceptable noise level. I'm not too sure why this is, but you could try the adaptive sampling patch and see if that helps at all.. And you needn't worry, BI isn't going anywhere any time soon. If BI works for you, then use BI ;) $\endgroup$
    – gandalf3
    May 19, 2015 at 8:24
  • $\begingroup$ That's a relief. BI still has its uses imho 8) . And I tried 500, and even 1024 samples, they look nearly the same as 50. XD $\endgroup$
    – Highstaker
    May 19, 2015 at 9:58
  • $\begingroup$ Could you post an image of 1024 samples? If it is still noisy you may have a problem. $\endgroup$ May 19, 2015 at 22:52
  • 1
    $\begingroup$ Yeah, maybe I was wrong about no difference, but even at 1024 the problem remains, the inner shadow is grainy. $\endgroup$
    – Highstaker
    May 20, 2015 at 4:55
  • 1
    $\begingroup$ I would comment only, but I don't have enough points. youtube.com/watch?v=h2UnvgqXqLc This is what you are looking for. It blurs the image based on input data. For your uses, normal pass will suffice. $\endgroup$
    – Alphisto
    May 20, 2015 at 11:17


You must log in to answer this question.