Different Works in Progress

Post your tests, experiments and unfinished renderings here.
User avatar
FarbigeWelt
Donor
Donor
Posts: 778
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: Kitchen, first preview

Post by FarbigeWelt » Mon Dec 10, 2018 8:47 pm

FarbigeWelt wrote:
Mon Dec 10, 2018 6:41 pm
Higher resolutions lead to better results with denoiser. 1xHD images was not nice.
I am pretty sure now that I know what lead to the disturbing low res images.
Instead of the bar light in the current scene I used two hollow spheres to get indirect light.
I may retry with the indirect light to be sure enough.
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+RX 5700 XT, 4/8 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Donor
Posts: 778
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Ocean in different Lights, or ring jumping was fun.

Post by FarbigeWelt » Mon Dec 10, 2018 10:02 pm

B.Y.O.B. wrote:
Mon Dec 10, 2018 7:28 pm
FarbigeWelt wrote:
Mon Dec 10, 2018 6:25 pm
Path rendrer seems not work with well with heterogenous, scattering medium.
I did not encounter any problems yet, all my smoke renders were done with Path OpenCL.
Maybe your problems were caused by something else, e.g. the spotlights being not-intersectable light sources?
BiDir lead so far to the only expected result. Most close path with three suns. However using point lights instead of spot lights improved the result remarkably. Now I have almost what I intended to get. Many thanks for the tip, and sure the dolphins would say "So long! And thanks for the fish!" (Douglas Adams)
Oecan in different lights, openCL path, BiDir.jpg
Oecan in different lights, openCL path, BiDir
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+RX 5700 XT, 4/8 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Donor
Posts: 778
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Denoiser and indirect lights

Post by FarbigeWelt » Mon Dec 10, 2018 10:17 pm

50 Samples in both cases, spot the different denoiser results!
Kitchen path 0.75HD denoised 2m18s, indirect lights, clamp 102593120.png
Kitchen path 0.75HD denoised 2m18s, indirect lights, clamp 102593120
Kitchen path 0.75HD denoised 2m18s, 102593120.png
Kitchen path 0.75HD denoised 2m18s, 102593120
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+RX 5700 XT, 4/8 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Donor
Posts: 778
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

House with 2 floors, each with 4 rooms

Post by FarbigeWelt » Tue Dec 11, 2018 9:23 am

At least I don’t have to care about statics.
The pool on to top 5*5*1.2 m^3 equals 30 tons or 1.2 ton per sqm.
117C8914-44E7-4ACA-9E4C-71CC0E6C0438.jpeg
Two floors House in the middle of nowhere
Last edited by FarbigeWelt on Tue Apr 09, 2019 9:40 pm, edited 1 time in total.
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+RX 5700 XT, 4/8 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Donor
Posts: 778
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Playing with Blender Particles

Post by FarbigeWelt » Wed Apr 03, 2019 9:12 am

Beer Foam 2.0

Background
Due to the fact my latest version of the dark beer scene „Hinech“ does not render foam through glass visibly anymore neither BiDir nor UniDir path, I started from scratch to get the picture I like to have.

Goals
Natural looking foam on (and in) liquids.
Faster particle simulation, shorter render times, reliability of scene setup.

Render times about 2 minutes for 1 Mpixel images on my 2 GPU.
Wireframe: Blender internal.

Funny Intermediates
BD6043FF-6EC4-46EA-97DD-B9216AB0327C.gif
Bubbles explode
BD6043FF-6EC4-46EA-97DD-B9216AB0327C.gif (1.78 MiB) Viewed 659 times
3D9489EB-29BB-44E2-8784-5920760246FD.gif
Wireframe
3D9489EB-29BB-44E2-8784-5920760246FD.gif (2 MiB) Viewed 659 times
0D14C398-E447-4960-9F19-0898522BCA22.gif
Dough growing out of box
0D14C398-E447-4960-9F19-0898522BCA22.gif (1.84 MiB) Viewed 659 times
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+RX 5700 XT, 4/8 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Donor
Posts: 778
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Particle Simulation in Blender

Post by FarbigeWelt » Fri Apr 05, 2019 11:11 am

After fooling myself with setting parameters of both Blender particle system and the plugin ‚Molecular script’ for hours, I found first steps for creating foam in a container.

The plugin simulates self collision between particles something what Blender unfortunately does not simulate.
As observed, there are many ways to crash Blender reproducibly with both simulators. Therefore auto-saving the scene every two minutes is a welcome feature if one does not like to get a very bad mood.

To use the plugin ‚Molecular script’ Blender Physics: must not be chosen ‚No’, otherwise particles keep their initial place, they keep initial place too if subframe is not set to zero.

Because particles jump if value .Timestep, in ‚Paricles’ submenu ,Physics’ is set to default 1/30 ([s] per frame).
Actually lower values make much sense, the particles’ animations get smoother and more accurate.
Lower values here mean e.g. 1/256 (observation: lower values than 1/1024 can crash Blender).

The plugin‘s simulation is visible, e.g. wireframe or solid, it starts after a click on the button ,Simulation‘.
This kind of simulation is slow because its speed depends on drawing speed, which is often at rates much below 25 [frames/s], especially if particles number larger than 250.

Default plugin value ‚mol_substep‘ is 4. This value can be decreased, although higher values increase simulation‘s accuracy, they also increase time. However, if value of ‚Timesteps‘ is e.g. a tenth of its default value; ‚mol_step‘ set to zero may fit as although results differ much using higher values due to resolving and then avoiding overlapping particles the better the higher the value of , mol_step is set.

Anyway, the simulation is much faster if started with a click on the button ‘Bake all Dynamics‘ in the ‚Particles’ submenu ‚Cache‘.
But way leads to different results because simulation is done with Blender ,Particles‘ and not with the plugin.

Currently, I am interested in understanding interaction of ,Particles‘ parameter with the plugin if only its self collision feature is enabled. After that I am going to enable plugin‘s particle linking feature. Sure I have already some experiences with that but it is very difficult to understand what its parameter do if there is an interaction with ,Particles, parameter.
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+RX 5700 XT, 4/8 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Donor
Posts: 778
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Four Colors Liquid Simulation: 64'000 particles

Post by FarbigeWelt » Sun Apr 07, 2019 10:33 pm

Code: Select all

Dimension: 2700 * 2700 pixels

Device: openCL (2 GPU)
VRAM: 880 MB
Engine: Path
Path Depths: T 42 (D 40, G 40, S 40)
Sampler: Sobol (adaptive strength: 0.60)
Light Strategy: Log Power
Filter: none
Denoiser: OIDN
PhotonGI Cache: default, no caustics
Clamping: 1e3

Export Time: 2.4s
Session Init Time: 15m35s  :?:  :!:  :?:  (mostly based on PGIC automat. lookup radius, see animated GIF)
Denoising Time: 14s
Manual Halt after: 6h20m13s
Samples: 239

Triangles: 86.0M 

Scene
 world, camera vol: clear IOR air
Lights: 4 (area)
Objects:
 1 tube (i vol: clear IOR glass, prio 5; o vol: clear IOR air, prio 0), 
 3 cuboid walls, 
 64k bubbles (subdivided 2 times; 2 layers; i vol: clear IOR water, prio 10; o vol: clear IOR air, prio 0)

Particle System
Systems: 4, 
Simulation: default plugin Molecular Script, 
Frames: 300, 
Time (baking): ~40m (~2s per mol_substep, total of 1200 mol_substeps)
Init-Render-4-col-fl-sim.gif
During Init Session of render, autom. lookup radius takes a lot of time
Four-Colors-Fluid-Simulation_1-to-216-samples.gif
During rendering (maybe path depth of 42 is bit too deep)
Four Colors Fluid Simulation in a Tube_GPU path_238 samples, 40 depth_dnpg no caustics_64'000 bubbles_6h20m13s.jpg
Picture this: Zoom in and Enjoy the the colors! Did you notice? Zoom-in leads to the impression of more saturated colors.
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+RX 5700 XT, 4/8 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
FarbigeWelt
Donor
Donor
Posts: 778
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Comparison Path depth 42 to 12

Post by FarbigeWelt » Mon Apr 08, 2019 8:47 pm

Four Colors Fluid Simulation in a Tube_GPU path_461 samples, 12 depth_dnpg no caustics_64'000 bubbles_11h32m53s.jpg
Four Colors Fluid Simulation in a Tube_GPU path_461 samples, 12 depth_dnpg no caustics_64'000 bubbles_11h32m53s
The picture above is from the same scene but with different path depths of total 12, the other three are 10 and the render time was 11h32 instead of 6h20m.

Brightness and gamma of following picture are heavily amplified to make the difference visible. With default brightness and gamma the picture is completely dark. The onion like rings are interesting, aren't they?
Four Colors Fluid Simulation in a Tube_difference 461 to 238 samples, path depth 12 to 42_gamma 9.9, brightness 10.jpg
Four Colors Fluid Simulation in a Tube_difference 461 to 238 samples, path depth 12 to 42_gamma 9.9, brightness 10
Indeed path depth of former render was not needed. But then, decreasing path depth did not alter speed much as one can see in the table below.

Code: Select all

time		Samples		mSamples/s
11:32:53	461		25.9
06:20:13	238		24.2
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+RX 5700 XT, 4/8 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

User avatar
B.Y.O.B.
Developer
Developer
Posts: 2809
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: Different Works in Progress

Post by B.Y.O.B. » Mon Apr 08, 2019 8:57 pm

FarbigeWelt wrote:
Mon Apr 08, 2019 8:47 pm
The onion like rings are interesting, aren't they?
My guess is: floating point precision limits ;)
Support LuxCoreRender project with salts and bounties

User avatar
FarbigeWelt
Donor
Donor
Posts: 778
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: Different Works in Progress

Post by FarbigeWelt » Tue Apr 09, 2019 9:16 pm

Four Colors Fluid Simulation in a Tube_GPU path_445 samples, 12 depth_dnpg caustics_64'000 bubbles_10h52m59s.jpg
Four Colors Fluid Simulation in a Tube_GPU path_445 samples, 12 depth_dnpg caustics_64'000 bubbles_10h52m59s
Four Colors Fluid Simulation in a Tube_difference 445 to 238 samples, pgi_with_caustics path_depth 10 to pgi_no_caustics path_depth 42; gamma 3.0, brightness 10.jpg
Four Colors Fluid Simulation in a Tube_difference 445 to 238 samples, pgi_with_caustics path_depth 10 to pgi_no_caustics path_depth 42; gamma 3.0, brightness 10

Code: Select all

time		Samples		mSamples/s
11:32:53	461		25.9
06:20:13	238		24.2
10:52:59	445		25.3
Behaviour with total path depth of only 6.
Four Samples.jpg
Four Samples
Six Samples.jpg
Six Samples
Thirteen Samples.jpg
Thirteen Samples
160.8 | 42.8 (10.7) Gfp / Windows 10 Pro, intel i7 4770K@3.5, 32 GB | AMD R9 290x+RX 5700 XT, 4/8 GB
17.3 | 19.0 ( 4.7) Gfp / macOS X 13.6, iMac 27'', 2010, intel i7 870@2.93, 24 GB | ATI Radeon HD 5750, 1 GB
#luxcorerender | Gfp = SFFT Gflops

Post Reply