Indigo Renderer Forum

Hi at all!
First of all sorry for my English, I hope it will be clear enough.

I'd like to share some tests made, with the same scene, under different Render Engines.
The scene in origin was taken from a videotutorial for Vray: it helps to build a simple studio for illumination tests and it contains a Plane\Background with rounded edges, a teapot, a text geometry and two planar emitters.

The target is to show (better that I can) which Render Engine generates the minimum amount of noise in a very short time (5min) with a resolution of 800x480 72dpi.
About tests made on "long distance" (hours and days of rendering), all the Unbiased Render Engines converge towards the same and little amount of noise and so the difference among them is very small; but since the time in this cases is really important I don't think that the only solution is to spend thousands of dollar for computers every year more and more powerful, but better to choose a software able to return a little amount of noise in a short time (few hours of rendering).

I insert at once what is, for my modest and shallow opinion, the Render Engines "ranking" consequent from the target above explained:


1 - Maxwell Render;
2 - Sunflow;
3 - Indigo;
4 - Luxrender;
5 - Kerkythea;
6 - Fryrender;



Obviously it's not simple to transpose the same identical scene on more Render Engines and try to optimize and set parameters for obtain a similar output, considering also that I haven't a decennary experience of these softwares.
Anyway I'll provide to insert this post on other forums regarding these softwares, and so I hope to receive, from more qualified people, the help necessary to optimize the scenes and change in case the ranking.

PC:
-AMD Athlon x2 4200+;
-2gb RAM;
-Windows XP pro SP2 64bit;

As I said the scene was set at first inside Vray and the output was taken as reference for the other Render Engines.

1) Vray v1.5sp1 64bit
Ris: 800x480; Time: 10min 27sec;
SCENE: http://www.ossosso.com/public/test/vray.rar
OUTPUT: http://www.ossosso.com/public/test/studio_Vray.tif

2) Maxwell Render v1.6.1 64bit
Ris: 800x480; Time: 5min;
SL: 8.18;
SCENE: http://www.ossosso.com/public/test/maxwell.rar
OUTPUT: http://www.ossosso.com/public/test/studio_maxwell.tif

4) Indigo v0.9.10 x64 + Maxigo v0.9.4
Ris: 800x480; Time: 5min;
SCENE: http://www.ossosso.com/public/test/indigo.rar
-Pathtrace + Bidirectional: Done 18630000.00000 samples - 60378.24375 samples / second
OUTPUT: http://www.ossosso.com/public/test/stud ... tional.tif
-MLT + Bidirectional: Done 19500000.00000 samples - 62387.99433 samples / second
OUTPUT: http://www.ossosso.com/public/test/stud ... tional.tif

3) Sunflow v0.07.2 + 3dsmax Exporter
Ris: 800x480; Time: 5min;
MinAA: -2; MaxAA: 0; Samples: 1; Filter: Mitchell
GI Engine: Path; Samples: 16
SCENA: http://www.ossosso.com/public/test/sunflow.rar
OUTPUT: http://www.ossosso.com/public/test/studio_sunflow.tif

5) Luxrender v0.1 RC4 64bit * UPDATE 24.01.2008
Ris: 800x480; Time: 5min;
Burn: 4.400
Environment: None
28.15 S/px
SCENE: http://www.ossosso.com/public/test/luxrender.rar
OUTPUT: http://www.ossosso.com/public/test/studio_luxrender.tif

6) Kerkythea 2008 Echo Edition 32bit
Ris: 800x480; Time: 5min;
Metropolis: rays 5/10000
SCENE: http://www.ossosso.com/public/test/kerkythea.rar UPDATED by Patrick
OUTPUT: http://www.ossosso.com/public/test/stud ... opolis.tif

7) Fryrender vBeta1.9 64bit
Ris: 800x480; Time: 5min;
Passes 7.5 :: 3.9 - Samples 11.55 M
SCENE: http://www.ossosso.com/public/test/fryrender.rar
OUTPUT: http://www.ossosso.com/public/test/stud ... output.tif
(values are a little different from Maxwell values to obtain a similar output)
OUTPUT: http://www.ossosso.com/public/test/stud ... meters.tif
(values are identical to Maxwell values)

8 ) Conclusions
http://www.ossosso.com/public/test/resume.tif

Naturally it's important the date of the post, that is 23/01/2007, in which Maxwell Render has a mature status of development, Fryrender is yet in beta versione and the rest are freeware.
This could in part (at the moment) justify "the ranking".

That's all.
Please, if anyone, after to have analyzed the scenes, would like to modify them to obtain the same output but with a smaller amount of noise in the same time, contact me and I'll update the post and, in case, "the ranking".

Thank you very much for your attention.

ossosso@libero.it

Nice of you to go through the pain of doing this.
I'm afraid your test is invalid as far as Indigo is concerned, though. As you can see, the image is completely out of focus, and it is the only render to be affected by this problem. In my experience, large unfocused areas increase render times quite a bit (in cases of images with shallow DOF for instance). You might still end up with the same results, but you would have to export the exact same scene to all renders (using Maxigo, you could feed the same 3DS Max scene into Maxwell, Fry and Indigo - don't know about the others' exporters...)

Hello ossosso, just some quick questions for you.
Anyway, I've seen some comparisons already. You should take a look at
them too! ... treddi.com, archi3t.it ... much better if you ask me. ;o))

edit: ... on the treddi forum you can also find a thread by isma
about the "Render Challenge v1" and the scene files ... click.

edit2: Ooops, I almost forgot Thomas An's Render Challenge ... click

And ohhh, you can also test Radium a bit, it's fun to play with Ian's toy!



take care
psor

Thanks ossosso !!
Very informative post.

Some remarks though. You are using wrong units in Maxigo. Your scene is modeled using meters, but your units are set to millimeters in Maxigo. That's the reason of the DOF effect. Please change the units to meters and re-render the scene. You may also want to turn on the Reinhard tone mapping, cause when using meters, your scene gets huge. Camera's focus distance is then over 130 meters. Reinhard would save you the time of re-tweaking the lights. Set Pre and Post values to 2 for a good start point.

Ps. There will be auto detection of Max units in newer Maxigo versions.

Thank you very much Suvakas for your advices.
The scene and the output were been fixed ad updated.

I tried to use "meter" under maxigo settings but effectivly lights didn't arrive to illuminate all the scene, so I passed to "millimeter".
Now using "reinhard" with a value of 2.6 for Pre and Post all works correctly.
Anyway noise amount remains the same.
Thank you again.

Psor:
1-I put the link to the vray scene, you can control there Vray settings. I used IrradianceMap+LightCache;
2-I didn't use DOF, and I didn't use Physical Camera;
3-Vray is an Biased Render Engine, it doens't make sense adjust settings to obtain a render in 5 minutes. Anyway you shouldn't worry about Vray. I put it in the post just because it was the start point of my scene, a sort of reference. It doesn't take part to the "challenge"
4-Because I haven't so many time of my life to spend in this way 1hour for test means 5 hours at all... well... also because if you wait too much, at the end all the result are very similar, in this way instead you can notice a big difference between Maxwell and Kerkythea for example.

@ossosso

Ok, I do understand your way of thinking ... know it comes ... but,
I recommend to use scenes that you would render later, and I guess
it will not be a teapot, heh?!

Anyway, you have to do what you like ... but I can't recommend the
way you're doing it, because it doesn't mean anything for production!

Sorry to sound blunt, but there are different reasons why I can't
support teapot renderings ... hehe. ;o))

We can talk about that later, if you can spend some more of your
precious life time ... !




take care
psor

Heya ossosso!

I now I bugged you already a lot! I just would like to show you isma's render
comparison ... I knew it was there just forgot about it. It would be nice if you
could try the test scene provided by isma to make your tests. For my taste,
his scene is way better! ;o))



"Time employed for the tests (20 minutes)
Used PC (amd dual opteron 265 - 4g ram - xp 32)"

Download Scene files:
-- OBJ + Textures
-- 3dsmax files

nb: Those tests are "pretty" old already (Aug. 2006), so if you like you can use
the scene files and do this test again! If you don't like, maybe we can get some
users to do this test again. I hope you don't mind ... ;o))




take care
psor