Method comparison

Feature requests, bug reports and related discussion
User avatar
Pibuz
1st Place 100
Posts: 2646
Joined: Tue Dec 11, 2007 7:58 am
Location: Padua, Italy
3D Software: SketchUp

Method comparison

Post by Pibuz » Fri Jun 17, 2011 3:43 am

Hi guys! After a few days, I had some time to compare the efficiency of the different render methods over the same test scene.

It is very simple: a plane with a box. Only one material applied: a phong plastic. Environment is a basic sun/sky system. I rendered all these images with SS1.

I wrote down some notes I want to keep in mind for myself. Maybe these may come in handy for developers too, who knows? :wink:

Generally, I find that the GPU acceleration gives excellent results (fast and with no artifacts) in the indirectly-lit areas, but it is still a little unefficient in the reflection areas (maybe it's a PT algorithm limit..). To get an overall accurate result it is still good to stick to PT+MLT, which gives correct results and it is faster than BiDir MLT.

IMPORTANT NOTE: I am running on winXP 32bits, so these tests are useful for the users who cannot locate large amounts of RAM, so they have to keep the supersampling to 1. Probably, increasing the SS to a higher number will produce less artifacts while using the PT algorithm.
Attachments
3.1.10 TEST.jpg

User avatar
pixie
Indigo 100
Posts: 2345
Joined: Sat Dec 29, 2007 4:54 am
Location: Away from paradise
3D Software: Cinema 4D
Contact:

Re: Method comparison

Post by pixie » Fri Jun 17, 2011 3:49 am

Nice!

User avatar
dcm
Posts: 663
Joined: Sun Jan 03, 2010 1:55 am

Re: Method comparison

Post by dcm » Fri Jun 17, 2011 4:21 am

thx but this so simple...try complex situation with small openings, double glass, metals etc.

User avatar
dakiru
Indigo 100
Posts: 747
Joined: Mon Jun 22, 2009 6:51 am

Re: Method comparison

Post by dakiru » Fri Jun 17, 2011 10:50 am

Hi, guys,
dcm wrote:thx but this so simple...try complex situation with small openings, double glass, metals etc.
I've made a simple test. Every image was rendered for 1 hour. Aperture diffraction is on and generated. MNCR is 10000, SS is 2. Two blackbody emitters from left (2500K) and right (8500K).

I was surprised a bit by the OpenCL result :P

edit: nVidia GeForce GTX 470, driver ver.: 275.33
01_Pathtracing.jpg
Pathtracing
02_Pathtracing_MLT.jpg
Pathtracing, MLT
03_Pathtracing_BiDir.jpg
Pathtracing, BiDir
04_Pathtracing_MLT_BiDir.jpg
Pathtracing, MLT, BiDir
05_Pathtracing_BiDir_GlAcc.jpg
Pathtracing, BiDir, Glass Acceleration
06_Pathtracing_MLT_BiDir_GlAcc.jpg
Pathtracing, MLT, BiDir, Glass Acceleration
07_Pathtracing_GPU_CUDA.jpg
Pathtracing GPU (CUDA)
07_Pathtracing_GPU_OpenCL.jpg
Pathtracing GPU (OpenCL)

User avatar
Pibuz
1st Place 100
Posts: 2646
Joined: Tue Dec 11, 2007 7:58 am
Location: Padua, Italy
3D Software: SketchUp

Re: Method comparison

Post by Pibuz » Fri Jun 17, 2011 7:08 pm

dcm wrote:thx but this so simple...try complex situation with small openings, double glass, metals etc.
Hey dcm!
These tests are only because I had a little spare time to use. :mrgreen:
BTW, if I spot some miscalculations or bugs in simple scenes, I assume they will appear on more complex scenes too; but in simple scenarios the developers can understand what's going on more easily, I think.

Thanks Dakiru for your tests! ..now we need some kind of point here: what are your final observations?

User avatar
Meelis
Posts: 383
Joined: Sat Mar 14, 2009 11:29 pm
Location: Estonia

Re: Method comparison

Post by Meelis » Tue Jun 21, 2011 5:56 am

Interesting, some are nice clean, some with noise and some show clear potential caustics.
What might be most accurate method :roll:

Alain
Posts: 101
Joined: Thu Aug 23, 2007 11:44 pm

Re: Method comparison

Post by Alain » Wed Jun 22, 2011 9:22 pm

I'm testing Indigo 3.0 as well for an interior scene.

Is there no way to let the sun come through glass and using GPU acceleration ?

Or the same question in other words:
With Pathtracing there is no way to let the sun shine through the glass of a windows ?


Kind regards
Alain

User avatar
Headroom
Indigo 100
Posts: 1061
Joined: Wed Aug 08, 2007 1:07 pm
Location: Spartanburg, SC, USA
Contact:

Re: Method comparison

Post by Headroom » Wed Jun 22, 2011 11:11 pm

Perhaps you are using the wrong glass. Tthe ghost glass from the MatDB is deprecated and should not be used anymore.

Glass should be modeled as a volume, not just a single plane. I am not sure the "Glass Acceleration" feature would help in your scene but it was brought to life particularly to speed up rendering for thin sheets of glass - windows essentially.

User avatar
PureSpider
Posts: 1459
Joined: Tue Apr 08, 2008 9:37 am
Location: Karlsruhe, BW, Germany
Contact:

Re: Method comparison

Post by PureSpider » Thu Jun 23, 2011 5:13 am

Also did a small test with pure PT with and without GPU, results attached.
Attachments
haus_gpu.jpg
CUDA enabled
haus.jpg
No GPU

Alain
Posts: 101
Joined: Thu Aug 23, 2007 11:44 pm

Re: Method comparison

Post by Alain » Thu Jun 23, 2011 7:43 pm

My windowsglass is a volume and I use an architecture glass material.

Pathtracing doesn't let sunlight throug glass, right ?

Where is the "Glass Acceleration" feature ?
I'm using Maxigo.


Kind regards
Alain

User avatar
Zom-B
1st Place 100
Posts: 4701
Joined: Tue Jul 04, 2006 4:18 pm
Location: ´'`\_(ò_Ó)_/´'`
Contact:

Re: Method comparison

Post by Zom-B » Thu Jun 23, 2011 10:30 pm

Alain wrote:My windowsglass is a volume and I use an architecture glass material.
The "architecture glass material" is a old man made fake material that nowadays brings more problems then it helps. The idea behind it is also to put it on a single sided window (normals facing inside!) not a volume.
Create a simple specular material with a IOR of 1.52 and transparency with uniform or white absorption spectrum. That should work like a charm!
Alain wrote:Pathtracing doesn't let sunlight through glass, right ?
wrong, but it may need longer ;)
Alain wrote:Where is the "Glass Acceleration" feature ?
I'm using Maxigo.
Should be to find under "Renderer Parameters"
polygonmanufaktur.de

User avatar
Headroom
Indigo 100
Posts: 1061
Joined: Wed Aug 08, 2007 1:07 pm
Location: Spartanburg, SC, USA
Contact:

Re: Method comparison

Post by Headroom » Thu Jun 23, 2011 10:31 pm

All sampling algoritms in indigo ultimately converge to the same result (at least in theory), so Pathtracing does let light through windows. The images in PureSpider's post right before yours are proof of that as both use Pathtracing (with and without GPU support).

Not sure what the problem is with your scene. Perhaps may want you post your scene in the Maxigo forum so someone can look at it and analyze it. Are you sure that the surface normals of your window volume point outwards and not into the glass volume ?

Alain
Posts: 101
Joined: Thu Aug 23, 2007 11:44 pm

Re: Method comparison

Post by Alain » Thu Jun 23, 2011 11:54 pm

@ Zom-B

When you say "specular material" you mean "glass/sss" or a "phong" material ? I can't find a specular material type in Maxigo.
I tested a phong material type called "Architectural Glass - Light" which was in the presetlist of Maxigo.

But both material types do not let sunlight throug the glass. It only works when I set the opacity to zero.

What else could be wrong ?
Im testing it with this scene:
Testszene_simpel_01_236011.zip
(46.87 KiB) Downloaded 201 times
I can't find any "Glass Acceleration" setting in the Maxigo render parameters nor in the Indigo Rendersettings.


Kind regards
Alain

User avatar
Zom-B
1st Place 100
Posts: 4701
Joined: Tue Jul 04, 2006 4:18 pm
Location: ´'`\_(ò_Ó)_/´'`
Contact:

Re: Method comparison

Post by Zom-B » Fri Jun 24, 2011 12:49 am

Alain wrote:When you say "specular material" you mean "glass/sss" or a "phong" material ? I can't find a specular material type in Maxigo.
I'm not a 3D MAX / Maxigo user, so I actually don't know the whole naming etc. but am versed in basic Indigo theory :)
It should be "glass/sss", but let the sss features disabled. The normals should point outwards of your glass mesh!
Alain wrote:I tested a phong material type called "Architectural Glass - Light" which was in the presetlist of Maxigo.
yes, as I told a old fake material... don't use it, it was nice in the "early days" of Indigo!
Alain wrote:I can't find any "Glass Acceleration" setting in the Maxigo render parameters nor in the Indigo Rendersettings.
Maybe you are not using the latest Maxigo version, it was integrated in Maxigo 2.0.6, but better use the latest Maxigo exporter from here: http://www.indigorenderer.com/forum/vie ... =9&t=10892
polygonmanufaktur.de

Alain
Posts: 101
Joined: Thu Aug 23, 2007 11:44 pm

Re: Method comparison

Post by Alain » Fri Jun 24, 2011 4:49 am

OK, sunlight still doesn't come in.

I use this material settings:
Screenshot.jpg
Result with Pathtracing:
Testszene_simpel_02_236011-scene.jpg
Result Bidirectionsal Pathtracing:
Testszene_simpel_01_236011-scene.jpg
The right half (red and black one) of the windows is just a plane, the normal points inward.
The left half is a flat cube. Normals pointing away from the cube.
Both use the material above.

I want the sun coming throug the windows like in the "Bidirectional Pathtracing" but with GPU Acceleration.

Is this possible yes or no ?


Kind regards
Alain

Post Reply
46 posts

Who is online

Users browsing this forum: No registered users and 45 guests