Test scences GPU Path vs CPU BiDir

Post your tests, experiments and unfinished renderings here.
Post Reply
User avatar
FarbigeWelt
Posts: 184
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

mix with null and equivalent opacity

Post by FarbigeWelt » Sat Jul 14, 2018 11:07 am

B.Y.O.B. wrote:
Sat Jul 14, 2018 9:33 am
I can reproduce a difference between mix with null and equivalent opacity.
It could be a bug, or maybe it is an expected behaviour, I'm not sure.
I think it would be good to log this in the LuxCore bug tracker, with an example scene.
Okay, I will open an issue today or tomorrow.
Microsoft Windows 10 Professional, intel i7 4770K, 32 GB, AMD R9 290x 4 GB, AMD R9 390x 8 GB
Instagrammer, please join #luxcorerender for your renderings.

User avatar
Racleborg
Posts: 124
Joined: Sat Apr 07, 2018 10:31 am
Location: UK

Re: Test scences GPU Path vs CPU BiDir

Post by Racleborg » Sat Jul 14, 2018 11:35 am

Hi FarbigeWelt

There were some good examples of different settings for the Velvet node, but the page is now down. I wonder if it would be possible to have the page available again?

It was here: http://www.luxrender.net/wiki/LuxRender ... als_Velvet

alpistinho
Posts: 1
Joined: Thu Jul 05, 2018 11:38 pm

Re: Test scences GPU Path vs CPU BiDir

Post by alpistinho » Sat Jul 14, 2018 12:10 pm

Hey,

The archive.org has a snapshot of this page. Maybe it could be used to recreate the equivalent page on the new wiki

https://web.archive.org/web/20170430090 ... als_Velvet

User avatar
FarbigeWelt
Posts: 184
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Scence: Opacity vs Mix with null Material

Post by FarbigeWelt » Sat Jul 14, 2018 9:59 pm

EDIT: Replaced with corrected version.
Opacity vs Mix with Null_2.zip
Opacity vs Mix with Null_2
(126.62 KiB) Downloaded 27 times
Microsoft Windows 10 Professional, intel i7 4770K, 32 GB, AMD R9 290x 4 GB, AMD R9 390x 8 GB
Instagrammer, please join #luxcorerender for your renderings.

User avatar
Racleborg
Posts: 124
Joined: Sat Apr 07, 2018 10:31 am
Location: UK

Re: Test scences GPU Path vs CPU BiDir

Post by Racleborg » Sun Jul 15, 2018 6:46 am

Allpistinho

Great, thanks for the info.

User avatar
FarbigeWelt
Posts: 184
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Fighting Fireflies

Post by FarbigeWelt » Mon Jul 16, 2018 5:42 am

My intenion here is to get a nice, wooden floor. It is using 7 different tiles based on the same material but different 3d mapping offset and slightly changed colors. Copied tiles and rotated them 180°, joined them and made some array copies. The walls have a bump map but you cannot see it.
The scene hower leads to many fireflies.
Wooden Floor Room with Spheres CPU BiDir Original.jpg
Wooden Floor Room with Spheres CPU BiDir Original
Any idea how to limit or avoid the fireflies? This render took around 5 h.
I'm using blender addon V2.1 alpha 2.

EDIT: add further pictures
Render Layers Composite.jpg
Render Layers Composite
Wooden Floor Room with Spheres CPU BiDir Despeckle.jpg
Wooden Floor Room with Spheres CPU BiDir Despeckle
Wooden Floor Room with Spheres CPU BiDir Denoise.jpg
Wooden Floor Room with Spheres CPU BiDir Denoise
Microsoft Windows 10 Professional, intel i7 4770K, 32 GB, AMD R9 290x 4 GB, AMD R9 390x 8 GB
Instagrammer, please join #luxcorerender for your renderings.

TryToRemember
Posts: 10
Joined: Tue Jul 03, 2018 3:35 am

Re: Test scences GPU Path vs CPU BiDir

Post by TryToRemember » Mon Jul 16, 2018 6:43 am

hey.

love the floor. a little too "new" or "perfect" but your method of making it be more unique is definitely clever!

So hey what are those glassy rectangles on the wall? 1-way mirrors to spy on the girls in yoga class? the world's shiniest blackboards? outside windows but with nothing existing outside? meant for HDRI to shine in? (why are they framed?)

Anyway is the specular on those things really cranked and are they dispersive? Turn them off or give them matte, and do a quick test... it's just a guess, but a guess from previous experience that screams "specular snow" (my term for this overall scene fireflies phenomena) . The distribution of the fireflies makes me think you have a world volume, with maybe homo/hetero scattering going on? And while I'm asking questions, why is everything in the scene smooth but the light tubes? Are you using a low-poly cylinder as an emitter and letting it be visible to emulate the glass envelope of a florescent light tube?

Also, you said "5h" but you know time and convergence, not really relevant (except to the poor human waiting on it!) too many variables, not the least of which is horsepower. Samples, on the other hand... how many spp was that?

Luxart
Posts: 41
Joined: Tue Feb 13, 2018 10:14 am

Re: Test scences GPU Path vs CPU BiDir

Post by Luxart » Mon Jul 16, 2018 6:53 am

Denoiser has remove fireflies option. It mostly cleans all the fire flies.
If you still get fireflies after enabling it, you can clamp the radiance using the clamp output option in render settings

User avatar
B.Y.O.B.
Developer
Posts: 1466
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: Test scences GPU Path vs CPU BiDir

Post by B.Y.O.B. » Mon Jul 16, 2018 8:22 am

In general, fireflies are unresolved caustics most of the time.
In this scene they probably stem from the metallic and/or specular materials.
The path engine offers a few more tools to get rid of fireflies than the bidir engine (e.g. limiting the indirect light visibility of the light sources, or limiting the length of glossy paths).
But if you want to stick to bidir, I think clamping is the best solution to get rid of them.

Edit: by the way, your light sources are very high poly. Consider using less triangles for them.
Currently you have 4536 light sources (emitting tris) in total.
Since they are all in one mesh, LuxCore will automatically switch off direct light sampling for those light sources (this happens with any emitting mesh with more than 256 triangles). This is done because usually, meshes with many triangles are not the main light source of the scene and would steal away computing power from the other lights, your scene is a special case.

I recently added the option to force the direct light sampling of meshlights to be on even with more triangles, your scene would benefit from this, however you'll have to wait for the next release to try it. viewtopic.php?f=5&t=9&p=4912#p4912
Until then, you should consider making your emitting cylinders more low poly, and keep them as separate objects so they stay under 256 triangles per mesh. (apply both array modifiers, enter edit mode, select all, P -> by loose parts)

Another thing I added recently is a warning if the scene contains many light sources. It is triggered in your scene and would probably helped you tell that something was not quite right :)
Attachments
scrn_2018-07-16_10-24-11.png
dls_off.jpg
(20sec) direct light sampling off (auto behaviour as in your scene)
dls_on.jpg
(20sec) direct light sampling forcibly enabled
Support LuxCoreRender project with salts and bounties

User avatar
FarbigeWelt
Posts: 184
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: Test scences GPU Path vs CPU BiDir

Post by FarbigeWelt » Mon Jul 16, 2018 10:52 am

TryToRemember wrote:
Mon Jul 16, 2018 6:43 am
So hey what are those glassy rectangles on the wall? 1-way mirrors to spy on the girls in yoga class? the world's shiniest blackboards? outside windows but with nothing existing outside? meant for HDRI to shine in? (why are they framed?)

Are you using a low-poly cylinder as an emitter and letting it be visible to emulate the glass envelope of a florescent light tube?

Samples, on the other hand... how many spp was that?

Thank you for the compliment. The floor‘s surface materiel is less ‚new‘ as it appears here.

The glassy rectangles are double glassed insulation windows. They are usually framed, e.g. painted wood for a good insulation.
It is night outside. I tried to use env. light but for some reason I failed with 2.1 alpha 2. Issue is opend, further investiogation is required.

No, as you can read from B.Y.O.B. And yes, are fluorescent light tubes with 6500 K.

Samples, only a few hundreds. But BiDir counts not the same way as path. Would be nice to have this information available after stopping the render.

I‘ve added the RGB gym balls for different reasons. The room was too empty and I’m still learning how to use glossy material. BiDir has been the choice bevause I expected a specific shadow. Path renders well and without much noise and fier flies.

And yes, I‘ve checked remove fire flies. Take the denoise picture and sharp it until you see some artifacts around them.

EDIT:12 h render added.
Wooden Floor Room with Spheres  2xHD CPU BiDir 299 S 12_08 h denoise 1100 s.jpg
Wooden Floor Room with Spheres 2xHD CPU BiDir 299 S 12_08 h denoise 1100 s
Last edited by FarbigeWelt on Mon Jul 16, 2018 10:28 pm, edited 1 time in total.
Microsoft Windows 10 Professional, intel i7 4770K, 32 GB, AMD R9 290x 4 GB, AMD R9 390x 8 GB
Instagrammer, please join #luxcorerender for your renderings.

Post Reply