[Bubbaloo]

And voila, here is the scene with a large emitter outside the window (with glass) again at S.L. 10:


Now I just need to get it to look like sunlight!

[oz42]

Interesting!

Were the times to reach sl10 for all these images similar?

With regard to making it look like sunlight; unless the sun is shining directly into the window what you really need to do is to make the emitter look more like skylight i.e. give it a bluish tint.

[Bubbaloo]

Times were approximately the same, around 8 or 9 minutes.

[lebbeus]

AGS/clipmapped/void materials are doing something funky with caustics even though they're not supposed to be creating caustics in the first place (remember my double-sphere test with the clipmap outer sphere and the 'butter' inner sphere??)

I can't agree because your example is only valid when there is SSS. Clipmaps and ghost layer should render in same efficiency like there-is-no-object without introducing anything funky.

Bubbaloo, can you try doing the following?
1) Remove windows and see if it makes a difference in terms of noise and I guess not.
2) Turn off/Hide emitters when rendering with sun and see if it renders much faster.

Tom, I have a thread that shows otherwise--I'll re-run things if you'd like

[Bubbaloo]

I'll agree that illumination is different when using AGS and no AGS/no glass geometry. The AGS is either somehow blocking some light or making the render take longer to clear/illuminate. Whether the resulting render is the same when cleared, I can't say.

[Tora_2097]

I usually pair a window shaped emitter with a ghost material applied to it together with physical sun. By being a ghost material the emitter will still let the direct sunlight and skycoor (important for reflections inside of the room) through will at the same time cares for a lot of the illumination. You will need multilight to combine them properly. I usually dial the sun down to about 10-20 percent so the noise level from the sun is at 10-20 percent only as well. You will loose some strength in direct lighting from the sun but the speed gain is dramatic, yet of course no longer physically correct.
(thats kind of my workaround for "portal lighting")

Benjamin

[oz42]

Tora - great idea and it obviously works because your renders are superb.

When you say that you 'pair' an emitter and a ghost material do you mean that the window panes have a single material with two layers; one emitter and one 'ghost' glass?

[Tora_2097]

not exactly. There are not 2 objects but a single one with a ghost BSDF (not glass, just complete transparent) and enabled emitter tab not 2 BSDF's.

Benjamin

[djflod]

well great workaround tora ... and like oz42 said the quality of your images is simply amazing.
Tora may I ask you what settings you have for your emitters ?

[tom]

Not much difference in noise amount. But the scene brightness is different.

I'll agree that illumination is different when using AGS and no AGS/no glass geometry. The AGS is either somehow blocking some light or making the render take longer to clear/illuminate. Whether the resulting render is the same when cleared, I can't say.

Now, correct. As you see AGS produce no refractive caustics. The difference is very natural as it seems you're forgetting the reflective layer on AGS. Of course, it would reflect some amount of light back but this has nothing to do with speed.

Tom, I have a thread that shows otherwise--I'll re-run things if you'd like

Yes, please let me know.

Ben, that's a nice workaround for compensating noise but I think you'll agree it's washing out the contrast, no?

[Bubbaloo]

The difference is very natural as it seems you're forgetting the reflective layer on AGS.

Yes, you're right.

The scene looks much better when using sunlight through AGS and allowed enough time to clear, but with this scene, optimization and speed is the key, forcing me to look at alternatives. Ben's idea looks like a winner. I will do more tests!

[def4d]

you could try to replace the physical sun by a big "laser light" setup?