LuxCoreRender Forums

Hi guys ,

I have this Project where i need to reproduce the effect of crystal material on
a lot of object. So I start experimenting with the shader node and engine settings.
And first it is impossible to have the Same look while rendering with Path and GPU/CPU.

Here are the realife reference :
Here are the Render with Luxcore bidir/path:
And The blend file

As you can see The Path version is completelly wrong . A bug ? And also The performance for this simple ball it Took 2Hrs on a i7 4930k.
Imagine that covering an entire Bathroom interior

Try to increase diffuse light path value. I don't know why, but changing diffuse light path value also influences on volume.

In the file i have uploaded diffuse path is already set to a very high value. Should be 64.
Look down to the first post.

Yes, sory. Now i see. It's really strange.
It's getting closer to bidir when i switch glass in architectural mode. But in this case we can not control rougtness.

Shadow color is making a difference. When set to black, all look ~same.


Edit1:
Scene modified to more natural size (sphere ~30cm, 1/10) and only difference is common - standard, minute difference between unidir & bidir engines (bidir has better light transport computation)

but I presume Dade would know, since he was already mentioning & fixing, planning things for future...

kintuX wrote: ↑Mon Sep 27, 2021 7:33 pm Shadow color is making a difference. When set to black, all look ~same.


Edit1:
Scene modified to more natural size (sphere ~30cm, 1/10) and only difference is common - standard, minute difference between unidir & bidir engines (bidir has better light transport computation)

but I presume Dade would know, since he was already mentioning & fixing, planning things for future...

Rivental wrote: ↑Mon Sep 27, 2021 7:18 pm Yes, sory. Now i see. It's really strange.
It's getting closer to bidir when i switch glass in architectural mode. But in this case we can not control rougtness.

Thanks to both of you. Really helping Idea. I tested them and they do improve the situation on GPU part. The only issue is that it remove all nice glowing and color blending of The Bidir+Realglass+Shadow color white Wich is quite close to the photo reference in the first post.

yeah, more tweaking

& Still advocating for an advanced glass (Smith model), microfacet nrmls, SPPM, computing spectral in wide color space...

Sharlybg wrote: ↑Mon Sep 27, 2021 3:36 pm As you can see The Path version is completelly wrong

You are using glass for the surface so the path is: entering glass (Specular) => scattering volume (Diffuse) => exiting glass (Specular). It is something path can render only in a very specific condition (if the path can easily than hit the light source).

Note: architectural glass does exactly work around the SDS problem to allow path tracing to render some SDS path.

Sharlybg wrote: ↑Mon Sep 27, 2021 3:36 pm And also The performance for this simple ball it Took 2Hrs on a i7 4930k.

It is an heterogeneous volume with multi-scattering enabled and 64 max. path bounces: as result paths will be extremely long.

You can disable multi-scattering or reduce max. path lengths to speed up the rendering. And/or increase heterogeneous volume step size and/or reduce the heterogeneous volume max. steps count to speedup ray marching inside the volume.

Reducing The bounces to 32 is the minimum i can set without loosing quality.And it help with perf a bit.
Below that it start to be too far from the reference. Architectural glass is a no go it look fake and unsaturated
And when you tweak the saturation on the material it start to look darker.

So finally only Bidir with deph 32 + White Shadow color + realglass + Multiscattering ON (OFF is too dark to match reality)
Will apply that to the project and see How it turn out

Thanks a lot !