oh, and THIS time, if you have any unwarranted criticism, follow this process first:
1)research global illumination for 1 year
2)write a paper where the thesis improves on a current GI algorithem
3)write a working demo in 3dGS with real-time radiosity (more than one bounce)
4)email me so I can stop wasting my time with my inferior coding

thanks in advance,
Michael Auerbach

more screens:

http://groundtacticsgame.com/gi_4.jpg
http://groundtacticsgame.com/gi_5.jpg
http://groundtacticsgame.com/gi_6.jpg

all real-time with 3dGS

-Michael-

Wow, a real 15 FPS screenshot..

It looks great, but:
May I ask who should use this shader tech in a reasonable timeframe? (in an actual game)

"Wow, a real 15 FPS screenshot"

actually, this is at least 3 times faster than any other method that supports both dynamic geometry and light positions. Real time radiosity lighting takes several times longer to render than direct-only lighting. No offense, but you obviously don't know what your talking about or you would already know this. I recomend you start your research here:

http://realtimeradiosity.com/

also, I honestly don't understand what your trying to ask. Do you mean how long before most commercial games use this technology? Probably a year or more depending on the application. Its not even used in 3d modeling/visualization programs yet which I find silly.

Also, most high performance gpus purchased today will run this at at least 40 fps with 2 bounces. On my machine, 1 indirect bounce renderes at 20 fps (the screens use 2 bounces).

-Michael-

Do you think with the tech talk a game will be more enjoyable?
Any tech reducing a simple scene to 15 FPS is simple not
suitable for a game, no matter if its
faster than other slow techs.

It simply possible to create better looking scenes
today that run much faster. For the player the details
are secondary.
You cant simply use the "high speed GFX Card" as justification.

Its good that you get into detail of shaders, I dont complain about how the scene looks,
but this is not suitable for games in the current techstatus.

"It simply possible to create better looking scenes
today that run much faster."

If its static, yes. This is dynamic, so, no.

Awesome job!

Is there any chance of implementing PRT (precomputed radiance transfer) into this? that's a feature i'v been wanting for ages... PRT is usually very fast compared to other realistic lighting methods.

to be honest I would be pefectly happy working on a commercial game where it is running an engine really nice looking at 15 fps, before serious per-game optimisation and a 2 year development cycle. By the time I finished, I'd expect 90% of my playerbase to see 60fps without a worry.

I love the idea of instant radiosity, even if its only on in indoor setting levels.

I just switched out half the models of the old beta to get the current polygons rendered on the scene to 860 000 polygons.

The engine was running at 8 fps with my 8800gtx OC. Visually, it looks not as good as crysis, even if the effects where exagurated for showing off what it can do, it wouldn't look as good as crysis with proper settings. Not at the current shape of the engine nor would it with the GI/Radiosity.

Performance wise it's not fast at all. 1 million and 600 000 polygons as you said a while ago, not gonna happen m8. I can run Crysis on very high on 30 fps and abit less on the snowy levels.

Besides this is without your "radiosity/GI", I would love to see it with it, 1 fps maybe?


and your quote "If its static, yes. This is dynamic, so, no."

Your GI/Radiosity is not looking good at all because of the low amount of bounces, nothing to do with your coding, it's only hardware limitation, and I don't expect to see half the market have good enough video cards to run this in 1 year or more. Crysis shipped on 2007, there are still ALOT of people struggling to run it at a decent framerate, now imagine your hardware heavy stuff.

as "the_mehmaster" said, it would be better to aim for PRT: http://www.youtube.com/watch?v=cC77J0xQbj4

I agree with the others, the presentation was not good enough, everything was exagurated and now this is the price you are paying for it. All the news post and technical stuff you said gave too much expectations in peoples eyes.

Might wanna think about that for the next time, never the less, it's a hardwork done and hopefully you will get the current state of it in a better shape in terms of visuals and speed.

//Endu

@foxfire: I really have to support damocles. Never forget.. A good game needs more than graphics. So if you take your pretty simple scene (15 fps), add some dynamic models (player, enemies - 10 fps), add BASIC Ai code (5 fps), add Basic collision detection (2 fps) this fps will drop, drop and drop.

So let's say it that way: Even if your computer is pretty slow (possibly?) it's just not acceptable if a Plugin crops the fps to the lowest value that's still playable.
Never forget the other very, very time consuming parts of a game (as already stated, a complex AI/collision/physics system might take a BIG bunch of your performance).

I do not want to critizise the screenshot itself, but if you claim that the way you've implemented the technique is the fastest way, than it simply not usable in any game that wants to be more than an graphical-demo.
A good graphic demo has to run with ~60 FPs per minimum to allow the code-Designer to implement the other parts of the game.

so final statement: Your method is looking good, but as long as it takes all the available performance it's not suitable for any Game.
greetings