LuxCoreRender Forums

I agree that the results of clamping in octane work way more predictable with little to worry about.
It's also interesting how you got very fast rendering even without rtx version, but I guess cuda always wins agaisnt opencl.

The PDF based clamping is an old JeanPhi's idea, he tried an implementation based on local vertex path clamping and it didn't worked well however I think the idea may work if considering the complete path PDF.

Can this be doable in the near futur ? want to battle Octane with right feature

About Clarisse IFX Clamping:

Since Clarisse 4.0 SP5, the path tracer offers a novel method to reduce considerably fireflies in renders. For more information please refer to the Fireflies Filtering section.

It's been developed in-house and no papers or other technicalities have been revealed yet.

About Octane GI Clamping (@ OTOY forums), I suspect it uses the method that Dade mentioned. Looks much alike or it's even the same to that used in VRay, ProRender, Corona...

I wonder if simply clamping only secondary bounces would help luxcore better.

And another though. Since photonGI isn't clamped at all, using low clamping with and without photonGI can give drastically different lighting. So if we could get clamping less affecting overal scene lighting we could also have more consistent results when using or not GI caching.

kintuX wrote: ↑Tue Mar 24, 2020 9:23 pm About Clarisse IFX Clamping:

Since Clarisse 4.0 SP5, the path tracer offers a novel method to reduce considerably fireflies in renders. For more information please refer to the Fireflies Filtering section.

It's been developed in-house and no papers or other technicalities have been revealed yet.

Maybe it's Antialasing Appearance in Global Illumination Renderers (2015).

kintuX wrote: ↑Fri Apr 10, 2020 10:30 am

kintuX wrote: ↑Tue Mar 24, 2020 9:23 pm About Clarisse IFX Clamping:

Since Clarisse 4.0 SP5, the path tracer offers a novel method to reduce considerably fireflies in renders. For more information please refer to the Fireflies Filtering section.

It's been developed in-house and no papers or other technicalities have been revealed yet.

Maybe it's Antialasing Appearance in Global Illumination Renderers (2015).

It is not related but this part:

"Furthermore, we show that covariance tracing can replace ray differentials as it removes its limitations (specular interaction) while maintaining backward compatibility (to easy integration). We believe that our method should in fine replace completely ray differentials (removing an unnecessary conversion cost)."

is very very interesting for other reasons.

is very very interesting for other reasons.

And you don't mind how hard i want to be one of your brain neural piece to understand that

Dade wrote: ↑Fri Apr 10, 2020 10:45 am It is not related but this part:

"Furthermore, we show that covariance tracing can replace ray differentials as it removes its limitations (specular interaction) while maintaining backward compatibility (to easy integration). We believe that our method should in fine replace completely ray differentials (removing an unnecessary conversion cost)."

is very very interesting for other reasons.

Haven't read it, but just watched the video and saw fireflies gone as a side effect... so, how do you find it beneficial otherwise?
Faster rendering - better quality, smaller mem. footprint, less fireflies & noise in less time from specular surfaces...? In short, will do.

Either way, I'm simply glad when you have an idea to further improve.

Stay well and healthy.

kintuX wrote: ↑Fri Apr 10, 2020 11:35 am Haven't read it, but just watched the video and saw fireflies gone as a side effect... so, how do you find it beneficial otherwise?

Image map caching (i.e. rendering with several terabytes of texture maps on disk and using the ram as a cache for mip maps and tiles). Ray differential is a requirement for this advanced feature but ray differentials have always been very cumbersome to support. If they propose an alternative, it could be very useful. Indeed, it would be a CPU-only feature.