pouët.net

3D acceleration kills the demoscene

category: offtopic [glöplog]
Seriously(?) though, I think that the gaming industry has to return to software rendering!!! =)))
It's a real power! How 'bout a realtime raytracing game? (Running at 2frames per second :)
Seriously(Again???) though, I will bring it back (calodox?) hehehe
Dalezy has to get my Ouzo backl again ;;;;;;;;0)
lolrotlfol
Seriously thouuughh,. I am thinking of a world domination project or social politics, to bring the software coding to the gaming industry and then 3d acceelarted carts (wow, wow, wow 3d accelerARTed! ART cards, wow!!! Hoh 0()_




Blah..
added on the 2003-05-13 09:20:39 by Optimus Optimus
oh, there is a realtime raytraced game. or atleast, its about to come up.

take a look at www.realstorm.com

I'm not kidding, thy FAN guys know how to raytrace :)
added on the 2003-05-13 10:28:13 by delta delta
and it sucks.

get real.
added on the 2003-05-13 11:09:16 by superplek superplek
Realtime raytracing is the future!
added on the 2003-05-13 11:45:19 by Optimus Optimus
nVidia just presented GeForce FX 5900 Ultra.
NVIDIA IS KILLING THE DEMOSCENE!
added on the 2003-05-13 12:14:32 by Scali Scali
I don't like bowling games either.

but hey. see it as a "proof of concept". RtRt games can be done. Gameplay is another topic here.
added on the 2003-05-13 13:25:37 by delta delta
I have ideas for realtime raytracing games, it's a pitty I can't motivate myself to start building my own raytracing engine ;P
added on the 2003-05-13 13:40:54 by Optimus Optimus
sure, one day companies like nVidia or ATI will make affordable for anyone functionnalities avialable in professionnal hardware like that:

http://www.saarcor.de/

so realtime raytracing is more than ever the future.
added on the 2003-05-13 20:09:08 by nystep nystep
Optimus: Don't you have ideas for everything in the universe ever, and too little time/motivation?
added on the 2003-05-14 01:10:46 by sagacity sagacity
Uninhibited sex kills the demoscene, or at least brings it to a temporary halt.
In an effort to turn this thread on to something worth talking about (this whole argument is completely retarded), I'm responding to the question posed by FooLman:

-Do all philosophers have an 's' in them?

...
no, not all philosophers have an "S" in them:

Georg Wilhelm Friedrich Hegel
Wilhelm Reich
Pierre Bayle
Ayn Rand [this is questionable - is Ayn Rand REALLY a philosopher? -- DISCUSS!]
Alan Turing [also: WAS HE A PHILOSOPHER?, or just a ponderous mathematician?]
Giordano Bruno

Just to name a few. Were you implying that only REAL philosophers have 's' in their names?... I have 2 in mine.
added on the 2003-05-14 04:53:13 by GltTcH GltTcH

KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL KILL THE FUCKIN SCENE ITS GETTIN TOO FAR!

added on the 2003-05-14 06:34:45 by raver raver
this thread ... kills the quakescene
added on the 2003-05-14 14:24:42 by uns3en_ uns3en_
Optimus: what kind of raytracing engine?

I have been reading a lot of applied scientific works which would enable me to write a replacement for a scanline renderer used in Blender, but OpenGl hardware-accelerated. Not realtime, but i intended it to be faster than software. I don't think it's posible on the leven of OpenGL 1.3 standard.

The major problem is, whereas the only restriction of software is slowness, hardware has more restrictios because it implements the most often used, minimal pipeline. You want something else - like one-component 32-bit txtures to implement depth-shadow masks -- you can't if it hasn't been foreseen. This changes now *a lot* with pixelshaders. Without them, you would waste away the whole hardware performance benefit or even much more when trying to circumvent the restrictions of the rendering pipeline, although it would be doable. For example, IRIX hardware verion of real soft shadow renderer turned out to be a few times slower than Pentium PRO software version. Go ahead and compare the performance of these computers when using a standard pipeline. But i believe a change is upcoming. When Ton Rosendaal was asked, what would be different if he wrote Blender now, he said, that the renderer would be OpenGL 2.0 instead of scanline.

-i.
added on the 2003-06-04 11:09:15 by eye eye
I have read about one of the raytracing cards and it was supposed to be taking control of OpenGL. I mean raytracing triangles instead of rendering them. But it uses the same elements. The type of raytracing I prefer is upon abstract geometric primitives, as seen in demos with spheres, cylinders, cones and again spheres, and objects cutting other objects, it's so abstract and generic and nicer looking perfect. Of course, raytracing on polygons is a good feauture too, if we want to construct simple polyhedra or cubic objects, but it would be preferable to use abstract objects for more circular things, like traycing on a sphere with a center and a radius instead of traycing upon the polygon mesh of a sphere. That's what I liked on raytracing engines actually and a only polygon raytracing would be not in my liking.

However, the main question is how to construct scenes with only these gemetric primitives? That's why polygons are needed too sometimes I guess. But the late FAN raytracing demos prooved to be able to produce more humanlike scenes by using geometrical primitives too. Their engine can trace polygons too though, but most of their objects are balls and cylinders and such cool stuff. Also, the games I had in my mind, were only based on spheres and stuff, very abstract worlds, simple arcade gameplay (something like Pang in 3dimensions and more). However, I had theoriticall ideas of how to trace diferrent objects than the regular ones without having to use polygon meshes. For me, that's on of the very diferrent and great properties of raytracing, that no polygon meshes are needed to be used (and there is also no z-buffering and other stuff) it's much more cleaner and smaller and generalized..

Ok,. shuting down ;P
added on the 2003-06-04 11:45:10 by Optimus Optimus
if the demoscene could be killed it would have been killed long before, but the demoscene is a spirit and there will always be people to carry on that spirt ...
the demoscene changed a lot, completely changed a few times over all those years ... that's the way life goes ...
added on the 2003-06-04 14:05:20 by poet poet
The late FAN demos could be reaplaced by combined OpanGL and software rendering, and most scenes by OpenGL only. There are 2 features which cause trouble: smooth shadows - which they don't use, and reflections on objects, which are neither flat nor round, esp. self-reflection. Though FAN did a great job, i somewhat believe that it could be done in a combined manner. Shadows are much more of a problem than reflections.
added on the 2003-06-04 19:07:05 by eye eye
midiclub : Yes, soft & hard are just some utils, every body can choose the utils he prefers/wants, and after we can criticize the prod, but not the utils like i was stupidely thinking when i've beginning this thread.
(when utils can be used for the same things i mean,
software rendering can make all the thing that hardware rendering and vice&versa, this is just a speed question... and please don't be disagree with this last sentence, pixels are able to create faces, and faces able to create pixels).
added on the 2003-06-04 19:36:14 by skarab skarab
midiclub: perhaps, but raytracing is more interesting and easier to do in software. How do you mean it though? Displaying something that look like raytracing or exactly the same but by using many polygons in OpenGL? Of course, someone can cheat and do many things even under a restricted hardware, but I don't like things that could be done pixel per pixel, to be simulated by polygons. For example, there were effects that were like oldschool 2d stuff, done under 3d accelerators, but they were more square like because they were using a grid of vertexes on X and Y and interpolation of colors in beetween, and no real pixel per pixel calculations! If OpenGL raytracing, would be to render an abstract sphere with polygons, then it's not raytracing. But anyways, I don't understand what you say. I was always fan of pixel per pixel output, especially for effects (2d stuff) and engines (rayrtacing, voxels) that are better to be done so and not by using 3d acceleration.
added on the 2003-06-05 12:42:28 by Optimus Optimus
1 polygon per pixel would be required for a genuine sphere, which would be kind of silly, especially if you had a scene with many spheres and the camera got close to the spheres... too many polies.

But you can get acceptable looking curves + spheres using beziers, nurbs, etc, which can be done with level of detail - then you always have good quality curves, without wasting billions of polys.

Shading can still be done per pixel, as can reflections in modern hardware. So you would not have too many problems...

Midiclub: FAN's engine does support soft shadows (through area lights i think but i may be wrong...) - take a look at the benchmark they made: http://www.realstorm.com/Download.html
added on the 2003-06-05 13:06:04 by psonice psonice
Optimus: Who cares if something's raytraced or not, as long as it looks good.
added on the 2003-06-05 20:49:54 by sagacity sagacity
midiclub, Skarab. I actually already tried combining hardware rendering and software rendering, in that same idea to combine raytracing and hardware triangles in the same scene.

However, the slowness to write to pixel buffer/zbuffer and the even worse annoying slowness to read them made it impossible in realtime. actually don't expect much more than 1 FPS even with a AGP 8x bus featuring a bandwith of almost 1Gb/s :-/

The only way to get some software rendered objects at decent speed with 3d acceleration is to use dynamic textures.

But forget about reading data from the video card, it's a very very bad idea.. ;-]

Most effect can be done with pixel shaders however nowadays. you'd better watch in that direction, or totally drop 3d acceleration, like FAN guys did.
added on the 2003-06-05 21:03:19 by nystep nystep
What did you use to grab the buffer? OpenGL's capabilities are clearly too slow. It appears to work better with rendering to DirectDraw bitmap. And i expect to get similar results from libSDL, though i have not tried. About 50 full frame reads per second seem to be a limit for GeForce 2MX of mine, as seen by some PSX emulator plug-ins, which have to grab the rendered image in most games at least in some cases.
added on the 2003-06-24 22:16:04 by eye eye
why are you guys all so dumb? stop coding, go collect stamps.
added on the 2003-06-24 23:44:01 by superplek superplek
what's the matter, plek? I believe that one has to search for better ways to utilise the hardware one may assume that is there. But to make new things it may also be requiered to resort to software or combine hardware and software rendering. The last seems to incur a huge penalty. If you think we are not good enough coders, give us tips how to code better. Else go collect stamps yourself.
added on the 2003-06-25 13:45:20 by eye eye

login