Raytracing GI Caustics, fast. (future):-)

I want all the accelerating stuff (pure,ice,targa3k) because you really CAN’T afford today to play with low scanline images anymore. The trend is going up to GI and caustics which you simply CAN’T simulate with a simple scanliner.

If you read the Interview of the Production Director of Final Fantasy, you will see how near the limits of pure scanline renderers like PRman are.

If the same models of Final Fantasy would have been rendered with GI and area lights, it would have been like a real movie.

The second reason (for me) is that the pictures need to get better and better, FAST. You can’t deny the real world enviroment, if you make photorealistic images. Beauty and subtle effects like reflection, refraction and GI will become sooner or later production standards.

Remembering the times where Pixar did the first rendered hollywood movie, Toy Story ? It was a technology jump. Now the industry enters the next major level, REAL photorealistic Images (the perfect lie :slight_smile: !

Entrophy is the only renderer so far, which is capable of GI COMBINED WITH THE FAST SCANLINE TECHNIQUES OF PRMAN, without loosing speed above scanline. But the idea of making hardware rendering chips, is above any speed barrier, and i don’t mean pure or renderdrive.

We are now like the 3d gamers were 10 years ago. There was no DirectX, there was no accelerated 3D chips, ONLY the CPU simulated the 3d enviroment (Doom,etc..).

Now, it’s the same for rendering, the cpu has to do hard work, for a thing that it was not developed for. IMHO, i think that EVERY professional task should be processed by special hardware, and that the cpu should only make the connections with the apps and the os.

Sure, the P4 has 2.0 ghz, but would you play a hi-end 3d game on software rendering ? O course NO !
No CPU can beats the speed of a geforce or radeon, it’s the same for hardware rendering chips too.
The cpu has no REAL 3d game capabilities, no matter how much they include MMX, 3Dnow! or SSE instructions.

Sorry guys, Prman is not out YET, but soon, if more hardware rendering chips companies appear at the horizont.
Renderdrive is cool, but it needs strong competitors for improving the performance and LOWERING the price, then you will see what hardware rendering can do :slight_smile: .
It’s only a question of time…

p.s: imagine, that someday you will play games in perfect quality, and doing rendering on the same 3D chip near realtime :-). (the geforces/radeons 4,5,6… are on their way :slight_smile:

What do you mean ? Tell me your opinion.

“If the same models of Final Fantasy would have been rendered with GI and area lights, it would have been like a real movie.”

I completely dissagree, i’d go so far as to say thats bull.
The problem with that movie was NOT the lighting.
using GI and area lights sure as hell would not have made
it look like a real movie.

The main problem was the motion of the characters.
The FF team was trying very hard to recreate reality,
and ended up being compared to reality(instead of being
compared to other CG flics). And all the characters moved
very bizarre compared to how a real actor moves and that crushed the illusion of reality.

besides, I think completely recreating reality is boring.
What do you get when you have a CG image that looks
exactly like a photograph?.. …just another photograph… whoopiee

-Tomi A.D.
Tomi@tad3d.com
http://www.tad3d.comhttp://www.tad3d.com

So what are you saying?

Are you saying that you are terribly dissapointed with the state of hardware technology.

P4/2Ghz too slow for ya, eh?

If it hadn’t been for the amazing increases in CPU speed over the last 10 years, we wouldn’t even HAVE GI or radiosity or raytracing at all.

Even the RenderDrive -the worlds fastest commercially available raytracing technology is only roughly 20 times faster than a P4/2Ghz.

Where does that get us for heavy scenes rendered with realistic lighting models?

Well let’s see…

instead of 20 hours per frame, now we’re at 30 minutes.

30 times faster than that, and we['re at a minute per frame.

60 times faster than that, and we’re now at one second per frame.

about 25 times faster than that, and now we have realtime Raytracing.

You do the math.

Unles somebody makes a quantum or DNA computer that you can afford, you’re [censored]-out-of-luck for years buddy.

kiwi

agreed…great way to sum it up.

…are you held accountable for crimes committed by your siamese twin?..

Ok so what your are demanding is that there be more and faster support for GI and caustics.. for whatever reason..

This isn’t entirely out of the question but the problem is that the hardware developer ends up creating a rather closed platform for one renderer. This makes developing unique shaders and the like difficult when you can’t compile it for the hardware.. because it wasn’t intially designed for it. On the other hand relatively inexpensive procs like the pentium or athlon will handle a wide range of software but is only optimized for mainstream applications. Unfortunately production 3D aniamtion is not very mainstream.. therefore the only accelleration we get is for games.

It would seem that with as many renderman compliant renders as there are currently that we could get some specialized hardware that would support GI and the like. It has been attempted before. I can’t remember what it was called buit there was a card for the Mac (eek) that made working with renderman doable. But these kinds of hardware allways seem to suffer from the continual upgrading of software and OS. OpenGL being a rather “open” descriptive language has made it possible for hardware developers to create really fantastic realtime performance.

However I think that what you hunger for is something that requires less optimization on the part of the artists. Realtime hardware performs well until you hit it’s threshold.. once it’s maxed out it comes to a screeching halt. A large part of 3D aniammtion production is also optimization and figuring out ways to avoid the long CPU intensive renders. Anyway my point being that it takes human intelligence to optimize.. maybe at some point AI and rendering will meet.. but I’m not going to hold my breath for that :wink:

Even if somebody perfected 3D so that it was really like shooting with film that doesn’t mean that suddenly every user will be a good director, competent actor or good storyteller.

Really a good story will overcome any shortcomings of the medium it was created with. Is Charlie Chaplin lame because his films were silent and shot in black and white?

It doesn’t seem like it’s too far off that the same display ahrdware that is used for games will be used for producing rendered images for broadcast and theatres. It’s still imaginable that film will be replaced by hi-rez projections. It makes sense when talking about 3D animation because it’s all synthetic .. you don’t really need a film projection. I’m sure somebody out there is working on hardware that deos GI and caustics in realtime but I wouldn’t expect it to be available and cheap in the near future. But if anything will driv ethe price down it will be it’s integration into games. Todays game system blow away what used to be state of the art mainframes that cost 100k or more.. os there’s progress. Wait about 5 years and maybe then you’ll have what you want.

g

I think, you did’t understand my post (except Gmask,partially).

  1. FF and the work of the artist on a 3d shoot, has NOTHING to do with the performance and fx (gi, caustics,etc.) !

How bad/good they did FF or what mistakes they made is not relevant for my questions in the post. Not to mention that if it would have rendered with gi, the FF team would have made shaders near reality. Yes, the motion capture and models were a bit unrealistic, but guys, it was the first movie of it’s kind.

  1. I don’t talk about the cost and/or the possibility (for now) to do that gi calculations fast. Much more there is important, HOW gi and/or caustics, subsurface scattering..(much more will come in the future) helps an Artist for creating images/movies of outstandig quality.

  2. Nobody will tell you that Gi is bull. I know that every pro who works on cgi combined movies, chooses scanline (prman) because of it’s speed an quality, FAKING gi,etc…
    Or would you tell me, that if you would have access to a superfast gi renderer with scanline-speed you would keep your shorts on :wink: ?

  3. Back to the roots ! Yes, games can be driven in realtime (what else:-), but why ? Because your P2 2Ghz renders it ?
    Gmask, you are not right if you say that you can’t make custom shaders for a HW device. Didn’t you see how Nvidia solved it for games (vertex and pixel shaders), beside THIS the Renderdrive ( I have asked them) is a specialized general purpose renderchip which can be updated later with a better rendering algorith and new functions (gi,caus,ssscatt.) VIA SOFTWARE.

There are some fact’s that can’t be denied:
Raytr.,GI,Caustics,SSS are indespensive for perfect images.
each game is GI (with backed gi textures) and more still images are gi made. Go on the softimage site or lightwave site to see the difference between. There are images where you can’t notice the difference between reality and gi/radios.

You will see, that like the superexpensive Ogl SGIstations were outperformed by a Nvidia Geforce in opengl, the same will happen with the next generation of renderers that will outperform an Renderdrive for only a few bucks for the consumer, or perhaps, the normal 3D cards will change devices with near realtime renderer capabilities.

DO YOU KNOW RENDERGL for 3DStudioMAX ?
It was made by Intergraph for raytracing and rendering on ordinary OpenGL chips :-). And it is FAST. It uses standard OpenGL calls to generate Aaliased full Raytracing. This standard is very old, BUT it shows that it can be made.

Believe me, IF someone would make an renderinterface like RenderGL with the whole Special programming found in the Geforce 3 (vertexSH,pixelSH), you could make a Geforce render engine.

P.s.: think about what you say, BEFORE you say it :slight_smile: !!!