Page 1 of 2

more tests

Posted: Thu Apr 14, 2005 2:38 pm
by daros
1.000.000 diffuse reflection polygons test.
Image

David Rossmann

www.arkadin.biz
www.animax.it
www.stack-studios.com

Posted: Thu Apr 14, 2005 6:59 pm
by Mihai
Daros, are you using the same method of chopping up the image into render regions and sending each region to a computer for rendering? Do you have to overlap the regions to hide boundaries or is that not necessary?

Posted: Thu Apr 14, 2005 7:12 pm
by tlass
Mihai if you split for example your image in four pieces,
my experience is that no overlapping is needed...

everything matches perfectly.... :wink:

Owerlapping

Posted: Thu Apr 14, 2005 7:59 pm
by daros
Yes, we have to render one pixel more for each side of the split.
The border pixel is darker.

David Rossmann

Test to simulate photometrics.

Posted: Thu Apr 14, 2005 8:13 pm
by daros
A total of 100 hours of calculation time for this image.
I try to obtain a Halogen-like photometric effect.
Each lamp has a 72 polygons reflector and a 6 polygons emitter.
The total is of 26 lights.
The result is still noisy.
Light bounces 8.
It is possible to archive the same result in shorter time?
David Rossmann

Image

Posted: Thu Apr 14, 2005 8:39 pm
by Maya69
i work the same subject

but what your computer

100hours is big !!!!!

what your number for polygone emiter ?

Interior

Posted: Thu Apr 14, 2005 8:54 pm
by daros
Hi,
26 lights with a simple cube for each emitter.

David Rossmann

Posted: Fri Apr 15, 2005 4:52 am
by aptoverde
hi can you post an image slice of the lamp, i think the effect you get is great... i dont speack ingles well...

Posted: Fri Apr 15, 2005 8:50 am
by Mihai
Is the material on the inside of the lights a metal, or plastic? Try putting a diffuse on the inside, make it almost white, and leave the metal material on the outside so it looks good.

Posted: Fri Apr 15, 2005 11:37 am
by Maya69
have try disable indirect reflection and refraction ?

We made something wrong.

Posted: Fri Apr 15, 2005 2:07 pm
by daros
I think we using the wrong approach.
We are splitting the renderings on multiple CPU using the render Region.
But for our test the Render Region is not optimal for Maxwell.
So, the rendering time i posted yesterday is not the result of 100 hours calculation of a single computer but one hour of 50 dual xeons.
The problem is that the result is much better on a single processor after 9 hours. For that i say that may the render region is not optimal for Maxwell at moment.

Sorry Mr Maxwell, we made something wrong and you are not so slow as i say yestrday.

you can se the results

9 hour on a single cpu
Image

1 hour on 50 cpu using Merged render region and not MXI files.
Image

Sorry again for the wrong information of yesterday.

David Rossmann

Posted: Fri Apr 15, 2005 2:52 pm
by Maya69
ouf i prefere

just a litle question

what your parametre for light fog

Light Fogh

Posted: Fri Apr 15, 2005 3:06 pm
by daros
I set the standard settings

Volume scattering, absorption 0.2, Scattering 0.2.
That's all.

David Rossmann

Posted: Fri Apr 15, 2005 3:57 pm
by Maya69
ok

i try with 0.2 both a fog parametre

but the fog is big

Posted: Fri Apr 15, 2005 8:51 pm
by oscarMaxwell
Hi all,

The split method to render one frame among different machines won't work for maxwell.
At the moment the only way is to render in cooperative mode under our net manager. In version 1.1.33 ( coming very soon ) the cooperative render have been improved considerably.

Best regards.