PhotonGI cache

Discussion related to the LuxCore functionality, implementations and API.
epilectrolytics
Donor
Donor
Posts: 790
Joined: Thu Oct 04, 2018 6:06 am

Re: PhotonGI cache

Post by epilectrolytics »

CodeHD wrote: Thu Feb 14, 2019 9:01 pm After all, LuxCore is advertised with a big orange "Physically based rendering" as the first thing you see on the homepage. That may raise expectations ;)
There is the rendering equation by J.Kajiya and I guess every algorithm that solves it without deviating can be called unbiased.
Pathtracers, including bidirectional ones, do that by design, but slowly.
With this the "physically based" thing is covered.
Basically every accelerator introduces some kind of bias.
Like Laci said, when you don't notice it, why not?

I think this new PhotonGI version is very interesting because you can control the bias.
Noticeable bias gets introduced by caching glossy materials.
But you don't have to cache those, you can exclude them and still enjoy the speed boost from cached diffuse which has only very little bias (there's not much visible difference to reference renders as shown various times earlier here).

But when you render technical stuff with mirrors and volumes maybe PGI will not be of much use and you'll have to resort to good ol' BiDir for which LuxCore is famous.

I think it's great to have all those options!
:D
CodeHD
Donor
Donor
Posts: 437
Joined: Tue Dec 11, 2018 12:38 pm
Location: Germany

Re: PhotonGI cache

Post by CodeHD »

lacilaci wrote: Thu Feb 14, 2019 9:12 pm Can you use a renderer and not be interested in visualisations?

"physically based" doesn't mean anything, same goes for "unbiased". Marketing terms, that's all...
Nothing is unbiased

But let's stay on topic...
My comment was not meant in a bad way, neither regarding the use of a cache in general nor the effort being done here.

It is possible to use a renderer without being intereseted in visualisations. I have used blender before to compare images of a straylight-rejection baffle to ray-tracing simualtions (from blender). It allowed me to easily check which surface-surface interactions caused the features seen and if it is a specular or diffuse effect. You can ask "why not use professional optics software like Zemax or CODE V", then the answer is partially the 10k€ price tag ;) I have seen other posts in this forum where people say they want to simulate lens systems, so a userbase definitely exists.

The only thing I want to understand (if only for myself) is, what sort of physical inaccuracies I can expect from certain settings - the cache in this case.
CodeHD
Donor
Donor
Posts: 437
Joined: Tue Dec 11, 2018 12:38 pm
Location: Germany

Re: PhotonGI cache

Post by CodeHD »

epilectrolytics wrote: Thu Feb 14, 2019 9:35 pm I think it's great to have all those options!
:D
Totally agreed, please don't see my other comments as being opposed to this ;)
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: PhotonGI cache

Post by Dade »

Sharlybg wrote: Thu Feb 14, 2019 9:30 pm So the only con of this cache is that you can't get boost if your average scene surface reflectivity is below 0.2 (wich is a default value that can be raise or lower down, but lowering it can produce heavy undesirable Bias)

So if you want to benefit a lot from it be carefull on the large surface area of you render ( floor/wall/ceilling/large furniture).
I'm going to change how this work for the obvious consideration than anything is better than a black surface inside the cache (i.e. I will use the material avg. over the hemisphere I was talking before, like I said, it will be anyway better than a black surface).

Outside the building process of the cache nothing will change: "glossiness" over 0.2 I will use cache, under I will use brute force path tracing.
Support LuxCoreRender project with salts and bounties
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: PhotonGI cache

Post by lacilaci »

Are you planning on doing some cache evaluation too? Like comparing neighboring samples etc, so that we can remove photon count and cache size and have some "quality/multiplier" parameter instead?

Also, do you have an idea on how to deal with those cache fireflies?
User avatar
Sharlybg
Donor
Donor
Posts: 3101
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: PhotonGI cache

Post by Sharlybg »

How many Time Do you need to match a PGI render noise level With PurePathTracer ? I made a comparison to spot how far Dade recent Development Skyrocket US. No OIDN in The mix just pure battle.





Hardware =======>> I7 8700k


PGI RENDER 1HR

PGI 1HR.jpg

PATH RENDER 9HR


Pure Path 9HR.jpg


9X Time Faster 9X Time Faster

To put things in perspective it's like Upgrading your workstation from a i7 8700k to 2X TR 2990WX (priced at 1800§ each = 3600§)
TR2990wx.jpg

The reason why i prefer by far invest in Software development than in Hardware :D


Only one things look like the light intensity issue is back (or my build is not the last one)
Support LuxCoreRender project with salts and bounties

Portfolio : https://www.behance.net/DRAVIA
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: PhotonGI cache

Post by lacilaci »

Sharlybg wrote: Fri Feb 15, 2019 7:41 am How many Time Do you need to match a PGI render noise level With PurePathTracer ? I made a comparison to spot how far Dade recent Development Skyrocket US. No OIDN in The mix just pure battle.





Hardware =======>> I7 8700k


PGI RENDER 1HR


PGI 1HR.jpg


PATH RENDER 9HR



Pure Path 9HR.jpg



9X Time Faster 9X Time Faster

To put things in perspective it's like Upgrading your workstation from a i7 8700k to 2X TR 2990WX (priced at 1800§ each = 3600§)

TR2990wx.jpg


The reason why i prefer by far invest in Software development than in Hardware :D


Only one things look like the light intensity issue is back (or my build is not the last one)
Great comparison!

Speedup is great but also with oidn you can get final renders even faster so in theory 2.2 is going to be even much faster for final results than 2.1 since you can now denoise better/and faster than with bcd. Not mentioning that since oidn is good at detail preservation, testing phase is much faster too so overall luxcore is now on par with the best renderers out there aaaand, if this gets on gpu it's gonna be the best!

Also regarding the brightness difference, I'm doing a lot of comparisons today against bidir+metro and you can get close with high depth values for pathtracing + high depth for photongi, but it's still somehow darker. Not sure what's up with that.
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: PhotonGI cache

Post by Dade »

lacilaci wrote: Fri Feb 15, 2019 6:41 am Also, do you have an idea on how to deal with those cache fireflies?
I'm not able to reproduce them anymore, do you have a test scene ?
Support LuxCoreRender project with salts and bounties
User avatar
Sharlybg
Donor
Donor
Posts: 3101
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: PhotonGI cache

Post by Sharlybg »

I'm not able to reproduce them anymore, do you have a test scene ?
Me too :?
Support LuxCoreRender project with salts and bounties

Portfolio : https://www.behance.net/DRAVIA
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: PhotonGI cache

Post by lacilaci »

Dade wrote: Fri Feb 15, 2019 10:35 am
lacilaci wrote: Fri Feb 15, 2019 6:41 am Also, do you have an idea on how to deal with those cache fireflies?
I'm not able to reproduce them anymore, do you have a test scene ?
I'm using build with recent russian roulette update for cache...

I get cache fireflies just as before, sometimes even with very hight numbers for photons and sadly sometimes even with large lookup.
Attachments
fireflies.jpg
Post Reply