GI Arnold Style

[Wegg]

How is that not as complex as any tree. . . or building. . . or car!?! Messiah kicks bootie.

[John Keates]

Nice test.

I assume that depth=reflection limit?

What is hammersley?

Is size=sample size (eg, you set it to 1 pixel)?

[chikega]

Hi John,
I recognize your name from the AM and Silo forums. :)

GI Depth is called Bounce in most other apps that I'm familiar with.

Hammersley is one of 10 Antialiasing Patterns in messiah along with 3 AA routines (Enhanced, Adaptive and Soft)

From the docs:
Size: This is a pixel measurement. The antialiasing determines the weight for how much the neighboring samples will be taken into account for the averaging with the original shaded point (center of pixel, or first sample taken). Usually that would only happen within a pixel (1.0), but this bigger radius (1.5 , 2.0) allows it to take samples outside of the pixel itself to get smoother results. This is particularly interesting for animations, since it could lower the amount of change from frame to frame and reduce buzzing. It acts more like a sophisticated "soft" filter but with less of the bad qualities of that.

[chikega]

Please bear in mind that I rendered the above with a three year old PIV 3.0Ghz 2GB Ram. Rush123 rendered the scene above with his 3.2Ghz Intel Duo 2GB RAM and it took him 8min 30sec. I need to get a new computer.:wip:

Just for grins I decided to rerender the scene using different Samples to see what effect it would have on render time and quality. I cropped the images to save on space and to zero in on where the noise is most prominent (in the shadows). The images read from left to right, top to bottom. All are rendered with GI Depth of 2.

GI Samples and time

2 1m40s
4 4m06s
6 4m23s
8 8m17s
9 8m12s
10 13m49s
12 13m59s <----8m30s on a faster rig
14 20m47s
16 28m41s

http://www.3ddmd.com/images/Messiah/PowerloaderSamples.jpg

[chikega]

I believe for this scene, GI Sample of 12 has reasonable quality for the shortest rendering time. I decided to rerender the scene by changing the AA Pixel Size from 1.0 to 1.5. This is similar to using the Soft button in LW which seems to have the same results as a .5 Pixel Gaussian blur in Photoshop. This is purportedly a smarter blur. It added one minute to the render time - 14min59s.

Size 1.0 and 1.5

http://www.3ddmd.com/images/Messiah/PowerloaderSoft.jpg

[John Keates]

Hi chikega, thanks for doing these tests and giving explanations.

My puter is not even as powerfull as yours but I am planing on changing that :) I think I can put up with some noise and add a little film grain to hide it. And the blur gives a photographic look too...

Outdoor scenes can be easier for a renderer to handle so I would be interested in seeing an in-door scene with lots of in-direct illumination with around 5 bounces.

It is late here in England at the mo and I have a busy next few days so I won't be able to try the demo for a bit. I see that it is save-disabled and only does low-res renders. I was kinda hoping that it would be a fully functional 30 day trial :)

[chikega]

Thanks John,
You know, using rather CPU intensive power features may not always be indicated. 5 bounces would certainly takes its toll. GI can still be faked by using carefully placed lights. The main feature you would look for in order to do this is the ability to exclude lights per object, to have the light not affect Specular and having negative lights - which most 3d apps can do including messiah.

I'm reminded of this XSI tutorial (http://www.3dtutorial.com/xsi.php?detail=25) of an indoor scene - it has the feel of GI but renders waay faster and looks cleaner - it's all done with carefully placed lights. Something to consider ...

http://www.3dtutorial.com/product_images/1.jpghttp://www.3dtutorial.com/product_images/2.jpghttp://www.3dtutorial.com/product_images/3.jpghttp://www.3dtutorial.com/product_images/4.jpg

[John Keates]

Thanks for the tips chikega.

You are right that I maybe shouldn't be expecting the computer to do that much... particularly as I am a painter really and can probably do pretty well by hand (given the scene to work from). I guess I just want to assure good results... get myself a good reputation and all that.

Actually, If I put figures in there then I will probably draw them by hand.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.