SSS differs according to render mode

[Polinalkrimizei]

So I need a cocktail and started with Whaat's Orange Juice Mat from the library.

I found that it works quite well in BiDir modes, but it turns out rather fishy with plain MLT or GPU accerlated PT.
Each image was rendered for 30 mins on an i7 2600k@4.6 Ghz with Indigo 3.4.4 and latest Blendigo, I exported the scene just once and changed the settings within Indigo.

Is there something I can do better to use my GPU for the cocktail or is it a known issue or what is going on there?

Cheers!

edit: I used a Kodak film to enhance colors/difference in color, so the OJ turns out very... orange. Don't blame Whaat.

[CTZn]

Both single MLT and GPU are based on PT !

[Polinalkrimizei]

Thanks CTZn, I just wasn't aware that PT and SSS are not friends...hardly use that feature.

[lycium]

That's strange, the GPU ought to reproduce SSS just fine... would it be possible for you to send us the scene (either as attachment or Dropbox link to support@indigorenderer.com please)? Normally single-directional PT with MLT is best for SSS.

BTW, "MLT" on its own doesn't specify a render mode, it is a toggleable option for the two path tracing modes (single- and bi-directional).

[Polinalkrimizei]

It's indeed strange as PT seems to work perfectly with my own SSS-juice... it just has an issue with Whaat's material. So for me the problem is solved, I attached you the file if you want to investigate.
Btw, did you mean .pigs or .blend?

For this scene, pure PT (GPU) seems best choice, second place goes to PT BiDir, much clearer than PT MLT!

[CTZn]

I am wondering wether there could be a misconception with the rgb scattering spectrum; it is not really defining a final color but rather the direction at wich the R, G and B components are scattered based on the incident ray. The scattering range is [-1.0; 1.0], ie, it is not a color but a "direction". Still, I am being approximative. To clarify the Indigo RGB editor does not allow for plain backwards subsurface scattering.

All in all the rgb scattering used seems to neutralize the absorbtion. If I use a pure white with a gain of 60, as compared with Whaat's uniform scattering, the render goes fine. That's no final answer, I'm just wandering about the issue

[OnoSendai]

It looks like it could be a modelling issue. The liquid mesh should overlap a bit with the glass mesh.

[Polinalkrimizei]

Thanks guys!

@CTZn: Your're probably right, I can't reproduce this with my own materials. Within Blendigo, I use a pure gain between 45 and 60, it works perfectly and it's consistent in all render modes.

@Ono: the "liquid must overlap with the glass! Liquid must overlap with the glass!"-mantra is sure always worth repeating. I did in this case, will post a bigger render later in the simple renderings thread if I have time, along with some ice an an orange.
With the new material, I am happy with the result, and pure PT (+GPU) works fastest and like a charm!