Page 1 of 3

High poly test

Posted: Wed May 07, 2008 10:45 am
by hafunui
It's somewhere around 152 million polygons, and 140 figures.

All instances, and rendered for 4h 45m. Still baking.

I'm thinking of adding more, and perhaps changing the materials.
_____

EDIT: added new version, now containing 1247 figures, and around 1.3Billion polygons. Thank god for instancing.

Time elapsed: 18 h, 7 m, 15 s
Done 372400000.00000 samples (727.34375 samples per pixel)
5708.56510 samples / second (175.17537 micro-seconds / sample)

Posted: Wed May 07, 2008 11:10 am
by fused
cool!

Posted: Wed May 07, 2008 2:22 pm
by alexmeyer
If you're gonna try a different material, you might as well make it a processor's nightmare and do SSS. :lol:

Posted: Wed May 07, 2008 8:12 pm
by Pibuz
+1 Alexmeyer!! :lol: :lol:

So, could you give us some pieces of informations about the camera settings, the scene lighting, etc?

Posted: Thu May 08, 2008 2:30 am
by Kram1032
even better, SSS with blends :D and a lot of indirect light :)

Nice test so far ;)

Posted: Thu May 08, 2008 4:03 am
by PureSpider
+1

Posted: Thu May 08, 2008 7:07 am
by hafunui
The very first render was made with glass. Even after an entire day, you couldn't make anything out. SSS would cause bloodshed.

Anyway, I've updated the first post with the new version, which is still baking. I probably could've added a lot more, but I don't think I could fit them in and make them visible at the same time :D

I wish there was an easier way to apply different shaders to instances. I cant do that in blender (to the best of my knowledge) so I'm assuming it's equally challenging with the exporter.

Posted: Thu May 08, 2008 7:15 am
by alexmeyer
:shock: :shock: That's one insane render. Very nice.


hafunui wrote:...SSS would cause bloodshed.
ROFLOL :lol:

Posted: Thu May 08, 2008 9:46 am
by Wedge
hafunui wrote:I wish there was an easier way to apply different shaders to instances. I cant do that in blender (to the best of my knowledge) so I'm assuming it's equally challenging with the exporter.
Yea, as a retired exporter developer for Blender/Indigo I can say this is not an easy thing to make work. :) Especially if you have multiple material instances that you want to alter in various ways. This, I think will need to be solved within Blender. (and also to my knowledge is not a feature of Blender, yet)

I suppose the workaround would be to set up a number of duplicates with different materials and then instance those.

Posted: Thu May 08, 2008 7:18 pm
by fused
thats quite insane :D

but can you beat this:

http://www.indigorenderer.com/joomla/in ... ewsIndex=1

~20.000 instances, each 7.2 million tris.
making a total of 144544154838 triangles (144.5 billion).

? :)

Posted: Thu May 08, 2008 8:10 pm
by BbB
Nice test. You should try to upload .jpg versions of your images. These are huge and take a long time to load.

Posted: Fri May 09, 2008 2:25 am
by hafunui
fused wrote:thats quite insane :D

but can you beat this:

http://www.indigorenderer.com/joomla/in ... ewsIndex=1

~20.000 instances, each 7.2 million tris.
making a total of 144544154838 triangles (144.5 billion).

? :)
Woah, what kind of render farm does he have. I guess it's worth a try :) But not with those scans; I haven't been able to get indigo to load one of those yet. (i need 64bit probably)

What material renders faster than diffuse? Phong? My images aren't getting any better than what you see above. Times like this make me wish I could bias it up a notch for cleaner renders. Maybe play with the bounce depths.

Also, my method of placement is use instances within blender using a rough hull which i model around the object, and with the center in the same spot (which is a fair ways below the figure) so Im not sure how to do that with a model I cannot load first into blender.

Posted: Fri May 09, 2008 2:38 am
by OnoSendai
Your albedos look rather high Hafanui, what are they?

Posted: Fri May 09, 2008 2:40 am
by dougal2
I think diffuse is not the fastest to converge if you're using a fairly high RGB value. There becomes much more reflecance = longer light paths = longer render times.

As for setting up an indigo scene with an external model - you could create a dummy model in blender, create instances of that and then manually edit the .igs to replace your dummy object with the external PLY mesh.
In this kind of .igs there should be a single <mesh> section for the "master" mesh and many <model> sections, one for each instance. You should be able to replace the single <mesh> section and all the <model>s will render that mesh. Just make sure you keep the <mesh> name the same.

Posted: Fri May 09, 2008 3:30 am
by hafunui
dougal2 wrote:I think diffuse is not the fastest to converge if you're using a fairly high RGB value. There becomes much more reflecance = longer light paths = longer render times.

As for setting up an indigo scene with an external model - you could create a dummy model in blender, create instances of that and then manually edit the .igs to replace your dummy object with the external PLY mesh.
In this kind of .igs there should be a single <mesh> section for the "master" mesh and many <model> sections, one for each instance. You should be able to replace the single <mesh> section and all the <model>s will render that mesh. Just make sure you keep the <mesh> name the same.
Yeah, thats how i did these. But how do I get the BB for the ones I cannot load into blender? Also, the object origins are way below the models (not in the center anyway). Would this affect the positioning in anyway?

Yeah, the bright shaders are pretty much pure white, I'll try with a lower intensity colour and see how that does.