pouët.net

Weasel : I just said that i wolud like to see demos that can do a slight fallback, but what i want is that compomashines show the demo in all the glory.. and maybe you se a slighly more dull version at home ( due to old hw)

Ryg : you can take this discussion as an input for a request to switch gfx-card.. or do we have to mail?

kb_ : I know that you can do alot with older techniques and software fallbacks.. why use dot3 when you can use EMBM with a unique texture per object, or why even that when Emboss bumpmap works on all cards? becourse ising the latest gives better result, and we want nice demos..

In dx you can talf about ps1.1 to ps2.0, but in opengl the only real pixelshaderstandard is ARB_Fragment_shader ( everything else i vendorspecific) and guess which chip is the only one that supports that and are availible right now? :)

but Scale is totally right, we arent really talking about ATI's radeon in generall here, were talking about dx9 class hw. but for now its just that ATI radeon is the only one supporting that.

MazyNoc: you misunderstood my point.

I wasn't talking about falling back to inferior techniques (like EMBM or emboss instead of dot3), I simply said that the things I've seen so far in GF3+ demos would all have been possible without a GF3, albeit slower.

Software VS are not really a "fallback", they're slower but they are 100% semantically identical. And as long nobody is using the texbeml/texdotXXX/texdepXX addressing modes or pixel shaders with more than effectively 4 combining stages, everything IS possible on a GF2 with dx7-style programming.

I don't advocate doing so, I'm SAD that all gf3+ demos I've seen could be done that way. As Xbox developer I know the way cool things which can be done even with ps/vs1.1, and with DX9 they're even almost doable on the pc ;). If I get to see a preview of a DX9/ps2.0/ARB_Fragment_shader demo which actually USES those resources and can't be implemented that easily using a GF4, I'll be happy to do my very best to convince Ryg that he wants a Radeon9700 in the compo PC or that we make an exception for that particular demo. And if the demo is doable on a GF4 I'll be happy to give advice on how to do that.

I have a 3dtexture effect ( can do it on gf2, but i need 3 times the texture memory, and it will not do filtering between layers)

Texture shadows.. (can be made on gf2 too, neds 2 tmus and 2 extra passes, and still only 16bit precision.. gf3 have 24bit precision in only 2 pass, and one tmu)

and I have plans for a envirnmental reflective voxel effect ( cubemap lookup from a 3dtexture that holds the normals in RGB) only gf4 has that texture lookup, but right now i only got a gf3 so that have to wait.. Right now its just Perpixel lighting on it ( needs 4 units or 2 passes, and ofcourse 3dtexture support, or use 3 times the memory)

yea, i know, its easy to just 'make up' effects to prove a point, but the fact remains, all the fun pixel stuff in opengl are vendorspecific except arb_env_combine - old but stable, and arb_fragment_program.. so the chices are, code totally different pathes for different cards ( im woking on that, but hey, i have a daytime job) or use the present HW's standard functions and let people se the effects when they get their hands on a newer gfxcard, or just settle with the older versions of the effects.

and for the speed.. we are able to push out an enourmos amount of polygons, and i know that it seams to be as many as we want.. but if we already have an effect that requires a huge amound of passes per frame ( 77 passes in one of my experimental effects) then a fallback to gf2 certainly will kill that effect, and it was 20-30 fps, even with all those passes)

People forgot the most important: You just can't push the limits on a PC ;P

Not even a serious text from me, hehe. And I haven't even written one quarter of this thread. Why are you talking about standards anyways?

Gotta stop messing with this thread and code a demo instead. Though, it will be on CPC..

written=read (fuck ME!)

:Gotta stop messing with this thread and code a demo instead

yeah, right 8]

it's a good thing we've got mindcandy so that I can watch the newer pc-demos.. pc-scene is for spoiled brats or rich godamn kids.. there's no chance I could get money to constantly upgrade my machine..

let's all go software, fuck hardware 3D!

kb: I don't really think that the argument of "it can be done on less hardware" is a very good one.
Sure, I agree that most demos don't use the hardware to the fullest potential... And that is a sad thing, totally contradicting the original spirit of the scene...

However, just because something can be done on less hardware, doesn't mean it's just as good, let alone better... Take Second Reality for example... It's done on PC, and on C64... But I still prefer the PC version... Higher resolution, better framerate, better music quality, etc, etc.

Anyway, that's not really the issue for me. The issue for me is that anyone writing nVidia-only demos should be shot. And a good way to get rid of nVidia-only demos is to stop using nVidia-hardware at parties. How about an R8500/9000/9100 then? (And ofcourse if people will start writing ATi-only demos then, they will be shot aswell, but we'll cross that bridge when we get there. First things first: get rid of nVidia's OpenGL-monopoly).

atleast you're having a good time.

this thread has been nominated for the "most useless thread at pouet" award.

if you have some cool DX9 demo, why not submit it in the wild competition? that's what it was ment to be, and i assume that the BP audience will be able to judge the difference between DX9 realtime and a VHS tape.

I don't think the demo-scene should be a motor to sell more expensive 3d-cards. I think demos released at parties should run on an average sceners pc in some way. the compo-pc was chosen by ryg with this in mind.

Its a huge difference between run on several gfx cards and beeing able to show its full potential.. and a compo should be able to view all effects of a modern demo..

If you submit it to wild, why not store it as an animation ( and beeing able to speed up the framerate a bit) and you can still brag about that all the effects are done by a certain gfxcard and could be realtime, even if its not really as fast in reality..

I can see why this thread is nomiated :-) .. mostly becourse ryg already stated that they wont change the demo HW ( well, i have to respect that, i cant do much about it), but look beyond BP for a while and think about this:

just how long should a gfxcard be availible until we can show demos on them? one year seams to be the normal delay.. so DX9 will become demoplatform in 2003, around the x-mas. Do we have the same limitations on CPU:s? ( then we can expect Amd 2400+ or less.. )

"ey, if you have a cool demo? why should you be able to compete in the democompo with it? comeon give the other guys with ugly crappy booring demos a chance! we want our german friends in smash designs to win with their really neat gf2 demos with ea sprites !!"

Scali: As I already said on this thread (2 times :), I will change the graphics card to Radeon8500/9000 *if there is demand for that*. And I also already said that Radeon9700 or a GeforceFX would be out of question simply because nearly no scener has it - Radeon8500/GF4 is already far on the upper end of what to expect of scener's PCs.

And I didn't say that I won't change the compopc hardware in any case, I said that I wouldn't change it "for such a bullshit argument". Which "we demand Radeon9700 because it's new and cool" is.

On CPUs: We'll choose something around 1.8GHz if possible.

And you don't have to mail to kb, all I want is a good reason to change the gfxcard to a Radeon.

Stefan: Yeah right, and ofcourse a Radeon9700 is going to make all demos goodlooking. Especially if a DXM demo is the only demo using it.

Fuck off.

it doesnt matter if you have a radeon9000 or Geforce4, the question was about dx9class HW or not..

The request for a dx9 class HW istnt becourse " it's new and cool", its becourse this is a major leap in how you make effects.. with floatingpoint buffers and fragmentprograms that doenst clamp each stage to [0..1] (well i know that ps1.4 can handle -2 to +2). This means that blendeffects wont look like 16 bits images as soon you blend more than one layer, and that you can calculate images woth much higher dynamic range and in a final step add gamma and thus gettting HDR like output.

ryg: i never said it will, but it gives them a chance.

why should i fuck off? you are the one not having a point and still talking bullshit here , not me.

iam not going to say "fuck off" to you, since its quite funny to listen to your attemps to explain your "point" :-D

"it doesnt matter if you have a radeon9000 or Geforce4, the question was about dx9class HW or not.." It does matter. And in fact that's the only question in the whole thread that ever stood to discussion.

I'm not going to repeat my argumentation again (which also isn't my lone opinion, as can also be seen on this thread). If you didn't get our (breakpoint organizing) intention at this point, you probably won't anyway.

"fr to huoston" *beep*
"huoston here over" *beep*
"we wanted to point out or point" *beep*
"point is noted loud and clear, thank you fr" *beep*

ryg: Consider me demanding Radeon (Parhelia is acceptable aswell).
Other than that... While Radeon9700 might not make all demos look better (which is debatable, since ATi has superior visual clarity and better and faster AA/AF than nVidia), it certainly won't make any demo look worse. So I don't think that this should be presented as a 'disadvantage'...

Other than that... About the CPU... Please use Intel Northwood... I will require large amounts of memory bandwidth/cache, and AMD CPUs just don't cut it. Clockspeed is not really an issue (although my initial target was 2 GHz, 1.8 GHz is no problem).

backing up kb/ryg on this. you guys are right. but stefan promised something that would push the nv20 to it's limits, (and as a xbox developer i know what it's capable of ;-)), so let's wait and find out!

see you guys there.

btw, as you allow dx9 in the demo/introcompo, and you only have DX8 hardware, isnt that kindof stupid? i mean why force users to download a hughe download?

but , then, can a dx8 demo use the size of dx9 as data files instead?

the point is:


there is this party - it has got its rules. go there and release your demo or let it be. but don't attempt to modify everything to fit your needs.


i remember people screaming years ago when ms97 set the .mod size to 500k, a few years later to 1.44mb and so on. as long as there are parties, there will be people trying to change the rules or competition computers. this is something one (who decides these things) has to be aware of and preferably not care about. anything else is useless. that's not related to arrogance or lack on social competence - it's just a waste of time cause you just can't cope with everybody's needs.


you should start to trust people who have already organized very good parties before.

The thread actually started, and went on with the debate about dx9 class hw or not.. "new dx9 demos wont run on their compomachines." as the first post said. And plenty of post points to that..
so its not really a ati/nvidia debate, its just that ATI are the only once having a dx9 card right now (or have you seen a gfFX in the stores near you yet?)