[bjorn.syse]

Actually the juiceglass i rendered earlier here which had a sss behind a thick glass did render well and clear noise pretty fast considering how difficult it would have been to render with 1.7.1 i think that sss in 2.0 is relatively fast if comparing to the old sss.

Yeah, you're right. That looked pretty good. Weird. Were you using the Liquid+Glass technique Thomas Ann have recommended?

[Maximus3D]

I think i used Thomas method, but i can't say for sure heh..
Btw, speaking of Thomas, i haven't seen him around on the forum for months. Where is he ?! his images always impresses me.

/ Max

[tom]

This particular scene is lit by Hyltoms LDR with white light from Top, Left and Back. I also added a quite large rectangular emitter to the right of the bottles.

But, it still can't save the SSS object from being illuminated by caustics only. As long is there is a dielectric between SSS and source of light, it will render slow. As you may agree it's nothing different than leaving SSS alone and attaching glass panels in front of your emitters.

- Avoid low attenuation and/or high scattering in relative to object's size.

Could you elaborate on this, I'm not sure I understand.,.

In principle, the light goes through the surface as much as you define by trans and attenuation. Then, it starts to decay and at the same time scattered inside the object. If your attenuation is low, it means you need more scattering to save the object appearing dark and noisy. Because, it means most of the light should be scattered back in the scene to make it appear decently. If your attenuation is high, it means the object is rather translucent and you should avoid high scattering, otherwise it will end up with bright dots or undesired coloring due to unrealistic entries. As it's not always possible to predict user's intention in making a material we try to avoid suggesting non-static range limitations on UI or in the core.

Ah, so this means any SSS liquid inside glass would take a lot longer to render, because all the light it deals with is caustic light, am I right?

Exactly.

Those tests are AMAZING! Could you say anything about rendertimes? I bet you're not using a Macbook pro though....

Thanks! These tests are done on a Dual Xeon E5430 and each take a couple of hours or less. I will try to share a scene to test on your side.

[RK_art]

Tom, thanks a lot for your tips.
Your renders look all amazing and right the way that I want mine too, but also after > 20 hrs I only got ugly grain within the SSS.
Good to hear that NL will take care especially about the SSS-rendertimes.

Yes, an optimized sample scene from your site as a template for our own work would help us a lot.

Don't get me wrong, I like MAxwell, otherwise I wouldn't have spend the money to upgrade all my licenses.
But the first results with SSS-speed , what was primarily my reason to buy the upgrade, were very, very frustrating.
I bought it for the speedup and saw..... almost nothing.

Adding some more optimized sample-scenes into the Upgrade-package could help to avoid these unpleasant experiences.

[tom]

Thanks, RK_art! Of course, we all wish SSS to be as fast as everyday materials but due to its nature and its complicated response to light it's the real enemy of simulation time. There are many ways faking translucent substances but they cannot go beyond offering a waxy look with no physical correctness so, we can't add these kind of cheap solutions. About V2 announcements and speed; it's never possible to guarantee a constant factor of speed for all materials and scene setups. Today, as you may agree, there's quite a big speedup in v2 for lots of other scenarios without sacrificing the quality/physical correctness and even pushing it more. If you've upgraded Maxwell expecting rendering SSS buddhas under sun much faster than you do in 1.7 and think the announcements were misleading, then I'd ask for some more insight at this point.

[polynurb]

Today, as you may agree, there's quite a big speedup in v2 for lots of other scenarios without sacrificing the quality/physical correctness and even pushing it more.

yes.. there are some scenarios unthinkable to render in 1.7

i did some complexity tests with the demo for the fun of it, using a totally hypothetical 2 layer weighted emitter carpaint shader including thin SSS...
(the blue one has no sss)

rendertimes vary 1- 2hrs, except one i left overnight.

It is great to see how the change in illumination strategy now seems to give much less noise at lower sampling levels.
these images looked pertty clean and had many details after 20 min already....

[bjorn.syse]

These tests are done on a Dual Xeon E5430 and each take a couple of hours or less. I will try to share a scene to test on your side.

Please do, I'd love to try out something optimized and compare the results and rendertimes. Btw, the Dual Xeon E5430, is that a Quad-core processor @ 2.6 GHz? If so, could one say that it should be approx twice as fast/effective as a Dual-core processor @ 2.6 GHz?

[kami]

funky images
they look great!

[RK_art]

If so, could one say that it should be approx twice as fast/effective as a Dual-core processor @ 2.6 GHz?

That depends on how your OS and the render-software can really use the separate cores and how they handle the available RAM etc.
Double the cores = Double the speed is only a theoretical calculation.

But I can say that Tom's tips really speed up V2's SSS-render-times a bit.

This one is 2 hours on a Quadcore 2,66 Ghz XP 64 system, only one big plane as Light-emitter, one SSS-Layer in the jade-material:



and the same after NeatImage-work:

[bjorn.syse]

Whoa! what a difference! So changing the lighting from sun to one emitter did the trick sort of? 2 hours is nothing, As I said, my image ran for 15 and was noisy as hell (probably because of caustics then). I'll actually see what happens if remove the PET-plastic bottling and just deal with the SSS-lumps.

Thanks for posting. That's a beatiful material! and very decent rendertime aswell! This gives me hope.

Tom, or anyone for that matter. Could I use some sort of workaround for my PET? Something like AGS that doesn't give caustics but provides some thickness and reflections in my case?

[macray]

Whoa! what a difference! So changing the lighting from sun to one emitter did the trick sort of? 2 hours is nothing, As I said, my image ran for 15 and was noisy as hell (probably because of caustics then). I'll actually see what happens if remove the PET-plastic bottling and just deal with the SSS-lumps.

Thanks for posting. That's a beatiful material! and very decent rendertime aswell! This gives me hope.

Tom, or anyone for that matter. Could I use some sort of workaround for my PET? Something like AGS that doesn't give caustics but provides some thickness and reflections in my case?

you could disable the direct and indirect caustics - maybe this helps as well?

[tom]

Please do, I'd love to try out something optimized and compare the results and rendertimes. Btw, the Dual Xeon E5430, is that a Quad-core processor @ 2.6 GHz? If so, could one say that it should be approx twice as fast/effective as a Dual-core processor @ 2.6 GHz?

Yes, it's 8 cores in total. As a tip, I can say it renders the benchwell scene approx in 9m30s with 1.7.

This one is 2 hours on a Quadcore 2,66 Ghz XP 64 system, only one big plane as Light-emitter, one SSS-Layer in the jade-material

Good to hear it started to be fun working with SSS. Btw, what's the size of your object? SSS is highly sensitive to wrong scene and object dimensions. It's also very fragile about colors and not so comfortable with arbitrarily altering them. I say this because it seems like you've manipulated the jade material a bit. Most of the time it should work but you should at least switch to HSV when picking colors and make sure you're not darkening or brightening them. Doing so would also need re-adjustments in attenuation distance and scattering coefficient.

Whoa! what a difference! So changing the lighting from sun to one emitter did the trick sort of?

Yes. Sun is like a distant/tiny emitter and the probability of rays hitting the surface is lower than other setups so the rendertimes go up. You can support emitters with low ambient such as skydome or a physical sky without sun, or better with IBL. That would make things even easier than you see now.

Tom, or anyone for that matter. Could I use some sort of workaround for my PET? Something like AGS that doesn't give caustics but provides some thickness and reflections in my case?

Exactly! AGS is quite suitable for PET bottles and its shadow won't be totally free of required brightness (see the AGS sofa in the recent post). But if you're still interested in making it dielectrics, matroska method will work.

[bjorn.syse]

Exactly! AGS is quite suitable for PET bottles and its shadow won't be totally free of required brightness (see the AGS sofa in the recent post). But if you're still interested in making it dielectrics, matroska method will work

Yeah, that AGS sofa was really impressive. So, fill me in, the matroska method is not the same as Thomas Ann's? Or is it a PET-bottle with wall thickness and then an SSS volume inside it just a small distance from the PET-wall, so rays pass through a little bit of air?

Would AGS still give any Attenuation effect? I reckon that's probably what makes PET look like PET, when you can see how it's a bit bluish or darker around the parts where it's thicker. If I were to use AGS, would I keep the geometry as it is?

Sorry for focusing this thread on SSS lately, perhaps we should have taken it in a separate one..

[tom]

So, fill me in, the matroska method is not the same as Thomas Ann's?

Yes, it's the only way to go.

Or is it a PET-bottle with wall thickness and then an SSS volume inside it just a small distance from the PET-wall, so rays pass through a little bit of air?

No, this is an absolute suicide.

Would AGS still give any Attenuation effect? I reckon that's probably what makes PET look like PET, when you can see how it's a bit bluish or darker around the parts where it's thicker. If I were to use AGS, would I keep the geometry as it is?

Yes, that's a problem. Although, you could try making the thick part apart at the top of the bottle made of dielectrics.

[micheloupatrick]

...Even if you have to, use Thomas's liquid+glass model as your modeling strategy...

Where can I find a description of this method?