PDA

View Full Version : Official Exluna statement regarding Nvidia


Array
08-19-2002, 03:09 AM
http://www.exluna.com/

"As partner of NVIDIA, we have an even greater opportunity to bring cinematics to games and interactivity to film production, and we look forward to working with all of you, our customers, partners, and supporters, to define the shape of things to come."

It brings quite a bit of relief to me that Exluna maintains commitment to film technology.

Grey
08-19-2002, 03:24 AM
But BMRT is still gone...

THANKYOU STEVE JOBBATHEHUT!

Array
08-19-2002, 03:50 AM
Well, it seems like Exluna is helping nvidia develop a GPU based rendering solution. remember what you heard in the Carmack keynote? Carmack stated that these card accelerated renderers will severly hurt software based renderers, some of which are using code as old as 15 years.

I think Larry Gritz will get his revenge on Pixar in one way or another.

Grey
08-19-2002, 03:55 AM
Yes indeed, I was thinking about that too!

Carmack's speach was a bit of a mind blower, but then he spoke about scanline render engines...

However, there are things I don't get with what he's saying.

Yes you can render scanline and get great results, but you have to do the same thing you do with Raytrace engines to get Radiosity simulation: LOTS of lights...

Then you have to activate shadows on ALL the lights in order to avoid what POSER users know as "The Glowing Nostril Effect"

I happen to know for fact this is true of all scanline render engines.

Now when you have large numbers of lights, then add shadows, you need one whole hell of a lot of computational power to get fast results...

I'm iffy on what Carmack said... some of it kinda sounds... well, maybe I'm a complette idiot... but he's talking about a mindblowing accomplishement either way....

Grey
08-19-2002, 03:58 AM
and BTW: I hope to god Larry gets his revenge... BIGTIME.

DotPainter
08-19-2002, 01:03 PM
Yes, this definitely bodes well for rendering performance.

Mr. Carmack is definitely on to something, but no one can say what the actual algorithms or techniques are that will be employed. It is still too early in the game *chuckle*. Most rendering algorithms are still built around limitations implied in software and general purpose cpu hardware. Each is designed around trying to simplify complexity and improve software performance while increasing the quality of render output. Rendering on GPUs with specialized hardware will require, IMO, rethinking the whole algorithm from scratch, in order to take advantage of the power of the gpu. Hardware rendering is not JUST taking rendering software and putting it on hardware.

Right now, they are just talking about using multiple passes through pixel pipelines to produce cinematic quality effects. Following this approach, the more powerful GPUs of the future will be able to process each pass quicker, thus allowing near real time rendering performance. I for one, don't think this is the optimal solution. I think that more powerful gpus should be doing more than rendering passes, but should also be optimizing processing to determine what,when and where to recalculate. Using things like dynamic radiance (radiosity) meshes, photon maps and lod algorithms etc, will be more useful to efficient generating hi quality real time output for everything: games, 3d software, movies and the like. That type of processing implies more tighter integration between 3d software and the gpu than a generic plugin, IMO.

It will be interesting to see how nvidia decides to incorporate their pipeline into 3d apps.....

kandyman
08-20-2002, 09:54 AM
Originally posted by Grey

Then you have to activate shadows on ALL the lights in order to avoid what POSER users know as "The Glowing Nostril Effect"

I happen to know for fact this is true of all scanline render engines.



Grey, not true. Not where The Carmack speaketh about cube maps influencing shading for a pixel. Facing the sky it gets bright and blue, facing the ground it gets red (in the case of Doom ]|[). I dunno about Poser, but JC sure has his shit together, and 500 passes for soft lights and maybe a bounce or 2 is nothing.

-marin

visualboo
08-20-2002, 02:41 PM
Originally posted by kandyman
I dunno about Poser, but JC sure has his shit together, and 500 passes for soft lights and maybe a bounce or 2 is nothing.

haha, no jokin brotha :)

Grey
08-20-2002, 03:31 PM
Kandyman, you sure about that?

I've seen this over and over again in several render engines... where light set not to cast shadow will shine through meshes onto other meshes and leave light anomolies that are a pain in the ass...

DotPainter
08-20-2002, 05:33 PM
Heres an interesting article I read on dynamic lighting in real time using dynamic meshing. Maybe John C. is against using raytracing in game engines, but it seems to me that variations of those algrorithms will be more promising than using simple lookup maps. Anyway, I can't dig into that too deeply, since I am not familiar with the concept. As the power of GPUs increases, I see a convergence of Gaming engines, 3d software and VFX software as they rely more on GPUs for render output.

Here one link:

www.graphics.cornell.edu/~parag/pair-preprint-medium.pdf

Here is an even more promising link

http://www.imm.dtu.dk/~khn/pubs/

kandyman
08-21-2002, 10:10 AM
Grey, say you render a Cubemap to determine lighting at a spot inside the nostril. The environment would be really dark except at some spot at the bottom where light could enter from e.g. a bright floor.

With this method, you don't need shadowmaps, as shadows are just the absence of light in this case. If you play around with modern GI renderers like Vray, you can see this effec nicely.

-marin

Grey
08-21-2002, 04:24 PM
kandyman, so no shadow maps at all... okay, I was completely unaware of that tidbit! Thankyou.

Black&White
08-22-2002, 07:48 AM
whats about real time GI.:scream:

Black&White
08-22-2002, 07:50 AM
waiting for real time Gi and area shodows...........

CGTalk Moderation
01-13-2006, 03:00 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.