GI Clamp

[Rudi]

Hi,


I have 3 images, each containing 1000 passages rendered by pmc time here.
As you can see the light in the image becomes less and less, lower the GI clamp is setting.

The first image with 1 million 1.000.000

The second picture with 5.

The third image I know with the GI clamp with 0.01

Honest said, somehow nothing real to begin.
Between the first picture with 1 Mill. and the other with 5 or 0.01 one sees in the light no real difference what concerns the rushing.
The pictures have been provided in a 4k resolution.

Greetings Rudi

[LudovicRouy]

This is not a fastest way of rendering,

Its to get cleaner render with less fireflies, for me it was very usefull on a bad hdri witch was useless until GI clamp, it keeps colors but with less power, there is far less "burned area".

http://render.otoy.com/forum/viewtopic. ... &start=420

[Rudi]

I had already read this, but had not understood.: (
My English is a translator.

My scene is with vulominösem light.
Basically the pictures in him become darker, the smaller the GI Clamp becomes. One sees less and less scattered light.


What I do not understand, is what you mean with HDRI. You mean this Painting? I use this only for a background lighting.

Can that not even explain to me one into German?

[KeeWe]

english:

As far as i understood the postings here, GI Clamp fixes some issues with fireflies and noise if using an HDRI (in LudovicRouy case) to light your scene but i think it also works with normal lights.
Obviously you are getting grainy results using an HDRI with a small resolution or, i don't now if this is possible, if they are in bad quality at all. Regarding your test: I think GI Clamp affects the intensity of your light distribution and IF you have materials which cause fireflies, GI Clamp reduces the chance of them. I don't now if its supposed to affect the overall appearence of your light setup. But i'm using Octane only since two weeks are still testing, trying and failing^^
If thats bullshit, please correct me. Don't wanna spread wrong information

deutsch:

So wie ich die Postings versteh, also die von LudovicRouy, führt GI Clamp vor allem bei schlechten HDRIs zu weniger Rauschen und fireflies (diese paar weissen Pixel die komplett ausbrennen). Ich nehm aber mal an, dass man damit auch normal belichtete Szenen schneller rauschfrei bekommt. Zumindest schlecht aufgelösten HDRIs bzw. HDRIs mit allgemein schlechter Quali (keine Ahnung ob man das auch noch an anderen Paramtern festmachen kann) führen zu mehr rauschen als HDRIs in guter Quali.
Was deinen Test angeht: Ich nehm mal an GI Clamp regelt die allgemeine Lichtverteilung bzw beeinflusst sie und wenn du eben Materialien hast die diese "hot pixel" verursachen, kannst mit anderen settings bei GI Clapm dagegen wirkt und zusätzlich schneller zu nem rauschfreien Bild kommen. Ich verwend Octane aber selbst erst seit 2 Wochen und kenn mich noch viel zu wenig damit aus. Falls ich grad totalen Blödsinn erzähl korrigiert mich bitte, nicht dass ich hier falsche Info unters Volk bring.

[LudovicRouy]

Abstrax explain it better :

Added new parameter "GI Clamp" to the path-tracing and the PMC kernels. It clamps the contribution for each path to the specified value.

By reducing the "GI clamp" value, you can reduce the amount of fireflies caused by sparse but very strongly contributing paths, i.e. it reduces noise by removing energy.

"Caustic blur", on the other hand, reduces noise by blurring caustics, but conserves energy.

That's why the more you clamp GI, the less energy lights have. But that keeps colors and that's why it works with my HDRI, the over way before that was to play with power and gamma but we loose colors.

[Rudi]

Now well as one sees in the pictures, reflections are swallowed bulky, the smaller the value GI Clamp becomes.
With daylight to pictures it probably plays no role, but pictures, quite darkly are become darker and darker.
The light dispersion decreases just.
I never had Fireflys with pmc.
Or are the Fireflys in the last picture? GI Clamp 0.01.

@LudovicRouy

How many passageways do you have the pictures in each case rendern leave?


@KeeWe
Danke schön

[inlifethrill]

Rudi, what is your question?

Generally, fireflies occur when a ray bounces more than 16-20 times once every million pixel sample. The rrprob parameter is meant to address the probability of this occurance. Another way to fight fireflies is by using gi_clamp. It's more of a brute force way that prevents rays from bouncing around for too long.
Yes, if pushed too far - brightness will be reduced as light bounces are reduced. It should be adjusted on per scene basis. It affects images rendered via the pathtracing and pmc kernels and works with any lighting scenario, not just hdri environments.

Hope this helps.
Dobs

[Augustronic]

To me it's very helpful to reduce noise and fireflies in complicated SSS-materials using daylight here.
I set it to about 10 and it's not reducing the overall brightness, only the artifacts.

[LudovicRouy]

@ Rudi, I think you ask for samples rate ?

The two pictures where cooked to 100 or 128 samples, but it's more than enough in my case to see the difference.

But the scene is very simple as it's just a head and a mixed HDRi for environement.