Test scences GPU Path vs CPU BiDir

Post your tests, experiments and unfinished renderings here.
User avatar
FarbigeWelt
Donor
Posts: 882
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Glossy Transluscent, correctum

Post by FarbigeWelt » Fri Oct 05, 2018 6:37 am

Sometimes a step back helps to fix issues.
I ran once again into 0815 issue, normals pointing into the wrong direction as it happens after common extrude.

Why glossy behaves differently regarding specularity if normals are wrong, but who knows?

Still, I think specularity is too good visible even set to 0.007. But then, sun is very bright and the behviour is pretty sure correct.
Glossy TL and Glossy comparison
Glossy TL and Glossy comparison
Test Joined Mesh simple.blend
Test Joined Mesh simple with corrected normals
(675.96 KiB) Downloaded 97 times
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
160.8 | 42.8 (10.7) Gfp | Windows 10 Pro, intel i7 4770K@3.5, 32 GB
2 AMD Radeon RX 5700 XT, 8 GB || Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Posts: 882
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Irregular Data Node, GPU Path vs CPU Path

Post by FarbigeWelt » Fri Oct 12, 2018 7:48 pm

Intended to help me understand DLC settings, I found a scene setup to visualize differences between CPU Path and GPU Path, not to mention their difference to CPU BiDir.

Instead of using Irregular Data node for Emission of a Matte object, I pluged it into Transparancy of a Glass object with a point lamp inside as base shape for LED bars.
CPU BiDir, CPU or GPU Path, LED with Irregular Data
CPU BiDir, CPU or GPU Path, LED with Irregular Data
Attachments
Test DLC.zip
Test DLC
(795.42 KiB) Downloaded 69 times
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
160.8 | 42.8 (10.7) Gfp | Windows 10 Pro, intel i7 4770K@3.5, 32 GB
2 AMD Radeon RX 5700 XT, 8 GB || Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Posts: 882
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Caustics from a Cube

Post by FarbigeWelt » Sun Oct 14, 2018 8:45 pm

Cube, CPU BiDir, depth 10, 100 Samples, 49m, denoised 0.25
Cube, CPU BiDir, depth 10, 100 Samples, 49m, denoised 0.25
Acutally I'm looking for settings for an ice cube.
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
160.8 | 42.8 (10.7) Gfp | Windows 10 Pro, intel i7 4770K@3.5, 32 GB
2 AMD Radeon RX 5700 XT, 8 GB || Gfp = SFFT Gflops

User avatar
Dade
Developer
Posts: 3369
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Irregular Data Node, GPU Path vs CPU Path

Post by Dade » Sun Oct 14, 2018 10:14 pm

FarbigeWelt wrote:
Fri Oct 12, 2018 7:48 pm
Intended to help me understand DLC settings, I found a scene setup to visualize differences between CPU Path and GPU Path, not to mention their difference to CPU BiDir.

Instead of using Irregular Data node for Emission of a Matte object, I pluged it into Transparancy of a Glass object with a point lamp inside as base shape for LED bars.
A path tracer is not able to transport a point light source across a specular (i.e. glass) surface. You have to use architectural glass. BiDir is able and it is explain the difference.
Support LuxCoreRender project with salts and bounties

User avatar
FarbigeWelt
Donor
Posts: 882
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: Irregular Data Node, GPU Path vs CPU Path

Post by FarbigeWelt » Tue Oct 16, 2018 10:28 am

Dade wrote:
Sun Oct 14, 2018 10:14 pm
A path tracer is not able to transport a point light source across a specular (i.e. glass) surface. You have to use architectural glass. BiDir is able and it is explain the difference.
Thanks for explanation, Dade.

But it does not explain why DLC circle artifacts are visible with GPU path but not with CPU path.
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
160.8 | 42.8 (10.7) Gfp | Windows 10 Pro, intel i7 4770K@3.5, 32 GB
2 AMD Radeon RX 5700 XT, 8 GB || Gfp = SFFT Gflops

User avatar
Dade
Developer
Posts: 3369
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Irregular Data Node, GPU Path vs CPU Path

Post by Dade » Tue Oct 16, 2018 11:43 am

FarbigeWelt wrote:
Tue Oct 16, 2018 10:28 am
But it does not explain why DLC circle artifacts are visible with GPU path but not with CPU path.
What circle artifacts ? In the posted image :?:
Support LuxCoreRender project with salts and bounties

User avatar
FarbigeWelt
Donor
Posts: 882
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: Irregular Data Node, GPU Path vs CPU Path

Post by FarbigeWelt » Tue Oct 16, 2018 1:21 pm

Dade wrote:
Tue Oct 16, 2018 11:43 am
FarbigeWelt wrote:
Tue Oct 16, 2018 10:28 am
But it does not explain why DLC circle artifacts are visible with GPU path but not with CPU path.
What circle artifacts ? In the posted image :?:
Please have a look at the pictures below Edit in:
viewtopic.php?f=5&t=430&p=6682#p6682
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
160.8 | 42.8 (10.7) Gfp | Windows 10 Pro, intel i7 4770K@3.5, 32 GB
2 AMD Radeon RX 5700 XT, 8 GB || Gfp = SFFT Gflops

User avatar
Dade
Developer
Posts: 3369
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Irregular Data Node, GPU Path vs CPU Path

Post by Dade » Thu Oct 18, 2018 8:56 am

FarbigeWelt wrote:
Tue Oct 16, 2018 1:21 pm
Dade wrote:
Tue Oct 16, 2018 11:43 am
FarbigeWelt wrote:
Tue Oct 16, 2018 10:28 am
But it does not explain why DLC circle artifacts are visible with GPU path but not with CPU path.
What circle artifacts ? In the posted image :?:
Please have a look at the pictures below Edit in:
viewtopic.php?f=5&t=430&p=6682#p6682
This is usually the result of some "incorrect" entry (like a light is not visible while it is) in the cache. It can happen if a light is hard to find and, with the default warmup period of only 12 samples, it ends to result always in shadow. The solution is to increase the "lightstrategy.entry.warmupsamples" parameter up to the point artifacts disappear.
Support LuxCoreRender project with salts and bounties

User avatar
FarbigeWelt
Donor
Posts: 882
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

CPU GPU Path

Post by FarbigeWelt » Mon Nov 12, 2018 7:34 am

Point Lamp, default w,r,g,b
Glass, rough 1.2
Text matte, emission: 1,10,5
CPU Path, Depth 8.jpg
CPU Path, Depth 8.jpg
GPU Path, Depth 8.jpg
GPU Path, Depth 8.jpg
GPU Path, CPU Path, Subtract, exp 10.jpg
GPU Path, CPU Path, Subtract, exp 10.jpg
Because of noise level and different distribution of noise (why, it is the same seed?) I smoothed the imgages with a gauss filter.
CPU Path, Depth 8, gauss 4 pix.jpg
CPU Path, Depth 8, gauss 4 pix.jpg
GPU Path, Depth 8. gauss 4 pix.jpg
GPU Path, Depth 8. gauss 4 pix.jpg
GPU Path, CPU Path, gauss 4 pix, Subtract.jpg
GPU Path, CPU Path, gauss 4 pix, Subtract.jpg
One can see there is not much difference between CPU and GPU path for this simple scene if its renders are smoothed.
GPU Path, CPU Path, gauss 4 pix, Subtract, exp 10.jpg
GPU Path, CPU Path, gauss 4 pix, Subtract, exp 10.jpg
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
160.8 | 42.8 (10.7) Gfp | Windows 10 Pro, intel i7 4770K@3.5, 32 GB
2 AMD Radeon RX 5700 XT, 8 GB || Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Posts: 882
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Noise, Path and BiDir or what are Samples?

Post by FarbigeWelt » Tue Dec 11, 2018 7:16 am

If I need 7000 path samples to achieve a similar result I get wth 100 BiDir samples, what is sample count for?
Three Monkeys and their Noise, openCL study
Three Monkeys and their Noise, openCL study
The Monkeys and their Noise, indirect only, clamped, HD, BiDir, 100 samples, 15min, BiDir one shot
The Monkeys and their Noise, indirect only, clamped, HD, BiDir, 100 samples, 15min, BiDir one shot
Indirect Light ,denoiser.zip
Indirect Light, denoiser
(235.13 KiB) Downloaded 71 times
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
160.8 | 42.8 (10.7) Gfp | Windows 10 Pro, intel i7 4770K@3.5, 32 GB
2 AMD Radeon RX 5700 XT, 8 GB || Gfp = SFFT Gflops

Post Reply