PDA

View Full Version : radiosity speed up


ital
07-22-2003, 10:44 PM
If there's one thing you speed up, please speed up radiosity two or three time faster, that's too slow, but i like too much the rendering,
Ital

aurora
07-22-2003, 11:17 PM
2 or 3 times faster!?! Man I would love to see the code that could triple the speed of radiosity that much! Not being mean or harsh or anything but after writting code for radiosity using the Monte-Carlo equ's I truely would love to see that kind of optimization even if it would require writting it in assembly. That would say nothing about how much I would love to use it if it ran that fast.

yog
07-22-2003, 11:26 PM
On the Newtek forum Lee Stranaham (Newtek spokesperson), said that there will be no significant improvements to the renderer in LW-8.0, although it's likely to be a priority in future revisions.

Here's (http://vbulletin.newtek.com/showthread.php?s=&threadid=8333) the thread here Lee mentions no "significant changes", along with possible reasons (his NOT Newtek's) why Modeler wont see as many new features as Layout.

yog
07-22-2003, 11:33 PM
Aurora, if you want to see 2 to 3 times faster improvements in GI rendering, then check out THIS (http://vbulletin.newtek.com/showthread.php?s=&threadid=8348) thread, where I lament that I do get at least 2 to 3 times faster (and often much more) GI rendering on other software.

True, most of the speed increase is in being able to concentrate GI samples into areas that need them most (unlike LW's blanket approach), and much better sample smoothing algorisums, but it's these sorts of optimisations that are needed.
Add in photon mapping to compliment the Monte-Carlo calculations and you have a much faster beastie.

Facial Deluxe
07-22-2003, 11:57 PM
Originally posted by yog
On the Newtek forum Lee Stranaham (Newtek spokesperson), said that there will be no significant improvements to the renderer in LW-8.0, although it's likely to be a priority in future revisions.

Here's (http://vbulletin.newtek.com/showthread.php?s=&threadid=8333) the thread here Lee mentions no "significant changes", along with possible reasons (his NOT Newtek's) why Modeler wont see as many new features as Layout.

... Rock'n Roll Thread for newtek....

aurora
07-23-2003, 12:19 AM
Yog, yeah I have seen that thread. I'm a lurker over at NewTek more then anything else. But like you said it comes at the cost of being pickey and making concessions on what/how radiosity will be applied. I would like to see a new alg for applying rad to entire scenes. I try to keep up with the material in the RadiosityBiblography but sometimes the math gets so intense I get headaches lol! Its true there are always ways to speed things up with different tricks/hacks. I was just talking about the over all alg itself. Although that link is a great one for people to learn from.

ngrava
07-23-2003, 01:28 AM
Have any of you guys see VRay for Max? I have it and it's seriously fast compared to LW's. Best of all, it's beautiful with shadows. A lot of the other GI renders can get sort of blotchy with little artifacts with shadows in the small cracks and details of your models. But VRay seems to work much better. I personally think it's better then either Brazil (cruddy if you ask me) and Final Render.

Just thought I'd mentions it. ;)

-=GB=-

Halsu
07-24-2003, 04:51 PM
The LW radiosity results can be made rather fast / decent with some optimizing AND enhanced AA + motion blur.

The settings can be kept really low (fast rendering), when the MB/AA passes blend them together.

Do NOT use HDRI as the BG illumination source. Using regular 24 bit image results in less grain, you should use real lights for the stronger, "main" lights anyway...

Of course, the fastest ways to render radiosity / GI look are to either bake the radiosity solution, or fake it, i.e. with spinning lights rigs - here's some example renders (http://www.kolumbus.fi/erkki.halkka/plugpak/PlugPak_Gallery_index.html) using the fake technique.

CGTalk Moderation
01-15-2006, 05:00 PM
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.