PhotonGI cache

[lacilaci]

Using latest daily I'm still seeing pretty big light difference and I cannot do anything to fix it

brute force with 5000 samples:

and a really dense cache with small radius:

[Dade]

Using latest daily I'm still seeing pretty big light difference and I cannot do anything to fix it

brute force with 5000 samples:
brute_force_5000_samples.jpg
and a really dense cache with small radius:
cache_dense.jpg

20 hours rendering Vs 2 hours ? It is still a good deal

Jokes aside, where can I find this scene ?

[lacilaci]

Using latest daily I'm still seeing pretty big light difference and I cannot do anything to fix it

brute force with 5000 samples:
brute_force_5000_samples.jpg
and a really dense cache with small radius:
cache_dense.jpg

20 hours rendering Vs 2 hours ? It is still a good deal

Jokes aside, where can I find this scene ?

It's actually 20 minutes vs 2 minutes yes, it is good deal and bias isn't too strong! Would be still great if the overall brightnes could be matched as close as possible...

scene is in your PM

[andreymd87]

It is nice indeed, but if you look at the wall light the cache is affecting lighting on the wall next to it, so I guess a little higher usage threshold would be needed in this case

using higher threshold will add more noise in glossy since it will be using more pathtracing calculation as far as i understood. so it's better to play with normal angle and radius

You would need very very small radius to make sure cache won't pickup the light at that place, which would also mean you'd need much more dense cache to get low enough bias, which will lead to long precomp time. Using higher threshold isn't as bad and it is going to make sure you don't get weird light leaks.

ofc. i agree.

[lacilaci]

I made small direct comparison on blenderartists: https://blenderartists.org/t/luxcoreren ... 1148794/15

[Dade]

Using latest daily I'm still seeing pretty big light difference and I cannot do anything to fix it

brute force with 5000 samples:
brute_force_5000_samples.jpg
and a really dense cache with small radius:
cache_dense.jpg

What puzzle me is that I'm quite convinced now that the problem is not in PhotonGI at all but in backward Vs.forward ray tracing results (aka Path tracing Vs. Light tracing). PhotonGI incidentally just does light tracing.

This is a rendering with PATHCPU (no PGI involved):

And this with LIGHTCPU (no PGI involved):

The difference is striking. I'm starting to think the problem may be in using "true" HDR and importance sampling (i.e. LIGHTCPU ends to emit only 1 out 1000 particles from low power HDR pixels).

[lacilaci]

Maybe the reason corona's cache gives exactly the same output as pathtracing is that they trace photons for cache from camera?(something vray does for lightcache too)

[Dade]

Maybe the reason corona's cache gives exactly the same output as pathtracing is that they trace photons for cache from camera?(something vray does for lightcache too)

The result must be the same (aside cases one forward/backward can not render) so there is something wrong (not related to PhotonGI as I wrote).

[zuljin3d]

Yes, there really is a difference in brightness.

HDRI, my old i5 intel cpu
PT vs PGI

[andreymd87]

ok. reverted the default settings of cache. but decreased a little the glossiness threshold. no light leaks or darker area. it's amazing. as i understand, the v2.2 will make me forget about corona or vray and 3ds max