Page 3 of 6
Posted: Sun Sep 03, 2006 12:31 pm
by Synthesizer
I just did a few quick tests and I'm really liking it. The tree cache option is great! It's so nice to not have to wait so long to see something.
Posted: Sun Sep 03, 2006 9:05 pm
by Behrendt
Ono-Sendai wrote:Behrendt: well, you don't *have* to use it
If you awnser that way, why do you start the poll?
Posted: Sun Sep 03, 2006 9:08 pm
by OnoSendai
If you awnser that way, why do you start the poll?
Heh.. just looking for an excuse really
Posted: Sun Sep 03, 2006 10:30 pm
by manitwo
Heh.. just looking for an excuse really
lol!
Posted: Mon Sep 04, 2006 2:44 am
by OnoSendai
BTW, there are some bugs to do with PDF calcs in 0.6 test 2 that i have now defeated, so take all your results/renders with a grain of salt!
Posted: Mon Sep 04, 2006 3:20 am
by psor
Ono-Sendai wrote:BTW, there are some bugs to do with PDF calcs in 0.6 test 2 that i have now defeated, so take all your results/renders with a grain of salt!
Well I'll call Adobe then ... *LOL* Sorry about my ignorance, what are
PDF calcs and what you do with it?
take care
psor
Posted: Mon Sep 04, 2006 3:23 am
by Kosmokrator
ooo yea PDF calculations of course....no one have idea what is this technical term...anyway i think is important!
Posted: Mon Sep 04, 2006 3:24 am
by OnoSendai
PDF = probability density function.
The bread and butter of Monte-Carlo ray tracing,
and really really really hard to get right for bi-dir tracing
Posted: Mon Sep 04, 2006 3:25 am
by psor
Thanx for the insight. ;o)
take care
psor
Posted: Mon Sep 04, 2006 4:06 am
by IanC
Grr lost my post.
http://apieceofstring.com/old.jpg
(warning 56k'ers 150k each)
http://apieceofstring.com/new.jpg
0.6 test1 vs test2. Both with 100 consec rejections, 20 mins each (amd64 3000 win32). All heads exactly the same mesh, 196,000 tris each. (Love the tree_cache!!)
Great work, Nick!
Ian
Posted: Mon Sep 04, 2006 4:12 am
by Zom-B
yeah... thanks for the insight!
I started testing a dimond scene, lighten up with HDR...
So here you got the 0.61 Render:
And here a 0.62 Render with enabled BiDir:
The v0.62 seems nicer for me... more "fresnel Style speculars" but color is killed...
I used a max_num_consec_rejections of 1000...
BTW:
Thanks Psor for the explanation of the Inifile.txt Stuff
Posted: Mon Sep 04, 2006 4:23 am
by manitwo
@IanC: really good test (very promising)
@Zom-B: i can't really see a difference except the color
nice scene btw
VERY GOOD WORK Ono-Sendai !!!
Posted: Mon Sep 04, 2006 4:31 am
by OnoSendai
Interesting test Ian,
hard to be sure of anything when there are bugs in 0.6test2 tho
Zom-B: very pretty, for once a HDR map that actually looks good!
Posted: Mon Sep 04, 2006 4:42 am
by IanC
Thanks Manitwo, the model has come along a little since then (shoulders) but I've finished with it, it was only a very quick test of the new blender features.
Nick, thanks
. The test2 version doesn't look much different to the test1 after a longer time, though the fresnel effect seems more accurate. Can't spot any smoothing issues, but then the mesh is high enough res it would be hard to.
I'm running a test with diffuse shaders now. Looks to be clearing up quickly.
Ian
Posted: Mon Sep 04, 2006 5:13 am
by Zom-B
He he... the HDR is a 8000x4000
HDR Example from Sachform (with watermarks!) but for (blurry) reflections its nice to use (you got to convert it to exr by yourself!).
Only Problem it totaly owns your RAM, in v062 even little more than v0.61... the 71 MB exr takes 376MB more RAM than Background lightning
btw. sometimes indigo needs very long to load this exr file... about 2 - 3MB in a second. While loading, my computer responds very slowly, but CPU time isn't taken more than 1%
In Network Rendering this happens (allways) too, but the exr loads much faster...
@manitwo
You have to save boath files and open them in your Picture Prog, skipping for and back to the other version, so you can see the differance much better
(hope anybody understands what I mean
)
