Denoiser CPU and GPU different results.

Use this forum for general user support and related questions.
Forum rules
Please upload a testscene that allows developers to reproduce the problem, and attach some images.
User avatar
FarbigeWelt
Donor
Donor
Posts: 1046
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: Denoiser CPU and GPU different results.

Post by FarbigeWelt »

Hi Dez!

Have you tried to compare same looking. not denoised images, by subtracting then, ouput file should have at least 16 bit. There should be a node for substract it in blender composer, maybe increase of exposure is required to see difference. If difference is good visible without exposure adjustment then you know the source for different denoiser results, hidden brightness if looking at 8 bit images.

CPU and GPU path show different results in many cases, e.g. light colors after passng through rough glass.
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
MacBook Air with M1
User avatar
B.Y.O.B.
Developer
Developer
Posts: 4146
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: Denoiser CPU and GPU different results.

Post by B.Y.O.B. »

FarbigeWelt wrote: Sat Nov 10, 2018 10:33 pm CPU and GPU path show different results in many cases, e.g. light colors after passng through rough glass.
It would be great if you could report these cases, because they usually are bugs.
User avatar
FarbigeWelt
Donor
Donor
Posts: 1046
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: Denoiser CPU and GPU different results.

Post by FarbigeWelt »

B.Y.O.B. wrote: Sat Nov 10, 2018 10:56 pm
FarbigeWelt wrote: Sat Nov 10, 2018 10:33 pm CPU and GPU path show different results in many cases, e.g. light colors after passng through rough glass.
It would be great if you could report these cases, because they usually are bugs.
Well, I did a while ago:

„Green light of LED bar (white point lamps, irregular data color input for glass' transparancy) is not visible with GPU path and only slightly visible with CPU Path but is overall visible with CPU BiDir.“
GPU path, CPU path with different green tone of LED
GPU path, CPU path with different green tone of LED
But could be fixed, I‘ll have to render the scene with beta 1 again.
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
MacBook Air with M1
User avatar
B.Y.O.B.
Developer
Developer
Posts: 4146
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: Denoiser CPU and GPU different results.

Post by B.Y.O.B. »

It would be best to have a much more simple scene, just the elements that cause the difference.
You can help us developers a lot by making test scenes as simple as possible, while still reproducing the problem.
And then it's best to open a github issue, posts in a long thread get lost very easily, while github issues can be tagged and searched much better.
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Denoiser CPU and GPU different results.

Post by Dade »

FarbigeWelt wrote: Mon Nov 12, 2018 12:33 pm „Green light of LED bar (white point lamps, irregular data color input for glass' transparancy) is not visible with GPU path and only slightly visible with CPU Path but is overall visible with CPU BiDir.“
As I have explained answering to that post: Path can not render lights beyond specular surfaces while BiDir can.
Support LuxCoreRender project with salts and bounties
User avatar
FarbigeWelt
Donor
Donor
Posts: 1046
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: Denoiser CPU and GPU different results.

Post by FarbigeWelt »

B.Y.O.B. wrote: Mon Nov 12, 2018 1:10 pm It would be best to have a much more simple scene, just the elements that cause the difference.
You can help us developers a lot by making test scenes as simple as possible, while still reproducing the problem.
And then it's best to open a github issue, posts in a long thread get lost very easily, while github issues can be tagged and searched much better.
Sincee Dade explained, I use architectural glass for path renders. In a short and rather simple scene I tested with glass rough CPU and GPU and found no issue in beta 1.
The bath room scene was a bit more complex because I found some strange effects using DLSC in more complex scene which was not reproducible with the simpler scene.

That is true, in github it is much easier to find entries.
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
MacBook Air with M1
User avatar
FarbigeWelt
Donor
Donor
Posts: 1046
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: Denoiser CPU and GPU different results.

Post by FarbigeWelt »

Dade wrote: Mon Nov 12, 2018 2:00 pm
FarbigeWelt wrote: Mon Nov 12, 2018 12:33 pm „Green light of LED bar (white point lamps, irregular data color input for glass' transparancy) is not visible with GPU path and only slightly visible with CPU Path but is overall visible with CPU BiDir.“
As I have explained answering to that post: Path can not render lights beyond specular surfaces while BiDir can.
I remember your explanation. My example here is a not a very appropriate one but the only one I have a record for. Still I think I observed different behaviour of CPU path and GPU path, but have to recall first and model them.
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
MacBook Air with M1
User avatar
B.Y.O.B.
Developer
Developer
Posts: 4146
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: Denoiser CPU and GPU different results.

Post by B.Y.O.B. »

Dade wrote: Mon Nov 12, 2018 2:00 pm As I have explained answering to that post: Path can not render lights beyond specular surfaces while BiDir can.
But that aside, shouldn't CPU and GPU path match each other?
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Denoiser CPU and GPU different results.

Post by Dade »

B.Y.O.B. wrote: Mon Nov 12, 2018 3:14 pm
Dade wrote: Mon Nov 12, 2018 2:00 pm As I have explained answering to that post: Path can not render lights beyond specular surfaces while BiDir can.
But that aside, shouldn't CPU and GPU path match each other?
Yes (but it is 150 samples Vs. 1500, hard to say with lights beyond glass and path tracing).
Support LuxCoreRender project with salts and bounties
Post Reply