Subsurface experiments

Oshyan · October 21, 2019, 02:16:03 PM

You have to right-click and Save As. Then you see the file name in the save dialog (you can click Cancel if you don't actually want to download it).

- Oshyan

archonforest · October 21, 2019, 02:28:33 PM

Impressive!

RogueNZ · October 21, 2019, 05:02:24 PM

Quote from: KyL on October 21, 2019, 08:49:07 AMSure. Render times are actually in the filenames of the pictures so if you open any of them in a new tab you will see the render time.
Usually it is between 30min and 1h to get a nice clean 2k image on my machine (threadripper 2950x with 32 threads)

In general I render using AA6, as most of the objects I used with subsurface are quite smooth and not very challenging to sample. I increase the number of path to 64 to get a nice clean subsurface. For animation you could probably go lower.

Thanks! I must be doing something wrong as a 1000x600 image of a displaced sphere with SSS, 0.7 detail and AA5 is taking nearly 4 hours (I'm using a 3900x). Will check when I get home

KyL · October 21, 2019, 06:48:17 PM

Oh. 0.7 detail is probably the reason. In my example in never went above 0.5

DocCharly65 · October 22, 2019, 01:25:49 AM

Very beautiful marble example!

RogueNZ · October 22, 2019, 01:49:57 AM

Ok I think my issue issue is the displacement (via PF) on the sphere. Turning off the displacement reduced render time from 3 hours to 7 mins. Is this expected? Rendering the displaced sphere without glass shader is very quick (minutes).

Another observation, using a sphere object, changing subsurface scatter in all directions (Beta) nearly tripled render time for no noticeable change in appearance. I know this was intended for use in thin/complex objects, but it's a worthwhile setting to check if actually necessary.

Dune · October 22, 2019, 02:05:06 AM

I did notice that adding another color to the density input increases render time too. This was the case in my ice experiment; bubbles only went faster than adding a constant color or PF. Which could be logical; more to calculate.

WAS · October 22, 2019, 02:39:59 AM

I wonder if mixing with a merge shader two separate glass shaders with their own colours with a mixer would help time

KyL · October 23, 2019, 01:18:05 AM

probably not, you would end up evaluating two glass shader instead of one, plus the merge operation.... $:-\$