pouët.net

Very true! I experienced this first-hand not too long ago. I had implemented a variation of the SCS algo for image-based CSG in D3D, and I had some people test it... One of them got 140 fps on a FX5200...
Then I had to tell him "Erm, I'm sorry, but I have a GF2GTS here which gets about 200 fps". And to put things in perspective... A Radeon 9600Pro got around 600 fps, and a 9800Pro was at 900, making the FX5200 an excellent target for an advanced point-and-laugh excercise.

It's all marketing. I bought a Geforce 4 Ti 4200 for 35E. And that because it was AGP 4X and not 8X.

The FX 5200 is a DX9 class card, it is AGP 8X, it has 128Mb memory, it's new, so on paper it looks great. And that's all that matters for marketing people and ignorant buyers.

in many cases, my GF2MX 32mb card beats my gf-fx5200 128mb performancewise. fx5200 sucks ass

for me, gffx5200 works better than my gf2mx did. and there are 3 factors driving me off from getting better card: NVidia has good Linux support and Kolor demos requiere them, so ATI is out of question, and 5200 is their first card since gf2mx which is cooled passively. i'm not a gamer, i'm trying to make myself a noiseless environment to work here. otherwise, it gives me possibility to test some shaders, and to do my usual modelling and animation.

BTW, now i found out that Zalmans heatpipe can cool gffx5600 passively, and i'm about to come up with even cheaper and more efficient passive cooling solution.

Hum... But Kolor demos don't work on linux do they? Well they are available for Windows at least.
And I can name a few good demos that will not run on FX cards, but will run on ATi cards, so what kind of criterion is that anyway?
And there are passively cooled ATi cards aswell, 9600 series for example (which are MUCH faster than an FX5200), and you can get 9800+ with Zalman aswell if you don't want to sacrifice performance.

Doesn't sound like your arguments are very solid.

Oh, and unlike the Kolor demos, the ATi-only (well they work on 6800 aswell) don't work on NV cards because NV cards simply don't support the required functionality, and not because the coders were too lame to write non-vendor-specific code.

i hope shiva wont read what scali said there :P

I'm going all-out with my new machine and gettting an X800. It'll be my first ATI card.

One Molex connector is enough!

Scali: that is a highly *personal* criterium, related tp my taste and preferances. If you don't understand such things, screw you, you asocial *thing*! And unlike those people who want to convince me that fx5200 is wrong, i don't want to convince anyway. as if i could.

Anyway, i'd like to know what demos requiere an ATI card.

I think you missed the point... The point was that you were comparing PS2 against FX5200, and said PS2 is faster than PCs.
So we pointed out that FX5200 is slower than some cards of many generations earlier, hence it is not very representative for modern day PC performance.
Then you started to give lousy reasons why buying an FX5200 would make sense.

who cares about boobs?

Scali: i re-read my posts very carefully, and i still cannot find myself saying "PS2 is faster than PCs". Well, it is damn unbelievably fast, considering it's a 1999 piece, but such a statement that it's generally faster than PCs would be plain stupid. (well, as if i never said stupid things anyway)

Anyway, i'm still waiting for someone to name me any Radeon-only demos.

Xbox is still better no matter what you say ;)

That was me, and I was ridiculizing him.

This is becoming a genuine The Obvious ((C) Scali) thread. Time to leave :)

Time to investigate abnormal maps, methinks.

my vote goes for paranormal maps

Yea, was it that obvious?
Was your SCS algo also much faster on gf2gts than on fx5200?

Whose *what* algorithm? :)

QED

plek: he probably means using stencil buffers to fake constructive solid geometry.

Congratulations, it only took you 2 days to google it.
And why do you use the word 'fake'? There's nothing fake about it, is there? You can render any CSG tree properly.

Scali: I'm assuming it took eye 2 days to google it because he has a life(*). Anyway, uruguay.

(*) Speculation, may not match truth

Scali: Because it's fake. CSG should produce a third object from two which you can edit/view/etc. later. Stenciling is just a quick way of displaying "something" like that.

csg-rendering is enough for many uses, so why not? one uses csg in raytracing without producing a third sphere or whatever either...

besides, when was the scene about anything else than faking?

I don't see how algos such as Goldfeather or SCS display '"something" like that'. They display exactly what the object should look like (unless there is something that you people are trying to refer to, yet fail to mention up to now?).
And the advantage is that there is no need to create new objects, so you can edit/view/etc in realtime, as opposed to other methods.

This is not about the scene btw, I don't think CSG could make a very interesting demo anyway. I developed the routine for work (CAD software).