hehe, concerning the nvidia/ati struggle:
i would prefer nvidia 'cause there are still 3dfx engineers working on the geforcefx and upcoming cards. but as you see, directx9 runs just better with ati and they're more innovative. not that nvidia's dx9 technology is worse, it's just that microsoft gave ati better cards than nvidia when they worked out the dx9 internal graphics chip architecture specifications. that's why nvidia's gforcefx series can't really perform to their full power! (that's internal graphics architecture, write if you want to know more about that) anyway, yes i like ati more than nvidia, they've just a better image + their cards deliver more power @ lower clock rates than nvidia (at the moment).
concerning the matrox parhelia: toooooo expensive, no dx9 support, bad game/driver support.
concerning xgi: it's all about the drivers, let's wait until they're good enough to release the card's full power (that's why most review are so different --> different driver versions).
concerning s3 chrome: that'll gonna be an opportunity! let's see!
anyway, most people here will have a new machine and an older one to run their voodoo cards and older games. i personally do not, but i should 
enough for today, more discussion tomorrow! 

I don't believe that any of 3dfx's old engineers are still employed by nvidia, or if there are, they're probably brainwashed into being average nvidiots by now anyways

Nvidia has shown in many cases that the only reason their cards can produce better benchmarks than ATI cards is due to drivers, which obviously points to driver "cheating". XGI has also become entangled in this practice, which is why I personally won't think much about getting a Volari until these issues are resolved. ATI on the other hand gets more power out of their cards. They also offer better low and middle-end cards, like my amazing Radeon 9500, which might not be the absolute top of the line, but still produces usable framerates with decent quality and AA/AF on, sometimes even all the way, depending on the game. On the other hand, GF4 MX can't even do half the crap necessary to play some games (Deus EX 2 comes to mind, although I'm not entirely sure), whereas a 9000/9200 are entirely capable.

Another point: when Nvidia tried to make Omega quit producing his third-party detonators, ATI commended him for his fine work with the cats (such a shame he's dropping 9x support  ).

But to counter all of this, I will throw in one good thing for nvidia: linux support. My GF2 GTS (a whole $10 investment) is now in my red hat 9 box, and nvidia's Linux forceware drivers are the only reason I can get decent gameplay out of TuxRacer 

Quote:


Driver cheating is the practice of lowering detail settings, adding extra optimizations, etc. for specific games/programs, normally 3D Mark, Aquamark, Quake 3, etc., programs that are normally used to benchmark the card. If you ever look at tom's hardware review, you will notice that almost all nvidia products accel at certain benchmarks, whereas ATIs are normally about equal in all benchmarks. ATI, IMOHO, focuses more on creating good products than lowering detail and making minor optimizations to make them look like they're good.

Quote:


What brings this up? Okay, YES. But in an entirely different way. MesaFX implements features via software that are not available on the card itself, which I suppose is cheating in a way.

Part #1
Thanks for your great definition amp_man, now I'm fully advised! Sorry, but it doesn't look like you know enough about today's graphics processor internal architecture to understand why nVidia "cheats" and why ATI is believed to do not. For your better comprehension - and inviting everybody else interested - here's a short summary of the problem you're all worried about:


We know ATI and nVidia to be the main developers for Microsoft when speaking of Pixel & Vertex Shading. Microsoft's task is to implement their work in the actual DirectX version, that's recently DirectX 9.0b. Due to decisions made by ATI and nVidia years ago, they gap between both firms' way of realizing new technologies became bigger. Along with DirectX9 a special topic became highly interesting, that one is, as you should know, internal arithmetic precision.
As the Radeon 8500-9200 (R2XX) works with a unique format (16bit precision), this is - simply said - not enough for complicated material shaders like the ones in DirectX9. So ATI gave the R300 and following models 24bit internal precision.
Meanwhile nVidia went another way: even the NV2X (GeForce 3/4) made use of 2 different formats: texture coordinats were calculated with 32bit precision, colour data got a mixture of 9bit and 10bit precision. To realize material shaders and the whole other implementations nVidia extended the colour precision to 12bit, and even a new 16bit format was introduced.
(Remember, it's also a matter if a GPU makes use of integer or floating point precision!
Integers: have a defined field of numbers and a fixed depth precision
Floating points: cover a wider range of numbers with the same quantity of bits but pay this achievment with a mostly lower, variable precision -> the higher the number, the lower the precision)
Anyway, Microsoft had to decide which format should be used by Pixel Shaders 2.0, and they grabbed 24bit precision. You see, if they would have set a higher standard, ATI wouldn't even be able to deliver DirectX9 hardware! But Microsoft also legalized a 16bit floating point precision.
See what that means? nVidia can't make use of their 12bit colour precision format combined with Pixel Shaders 2.0! Here's why this fact hits the GeForceFX series so hard: the NV30 has 2 arithmetic units: the first handles Pixel instructions with 32bit. This high precision hits performance -> means that most operations could only be done 4 times per clock. That'S a bit too poor, so nVidia decided to create a second unit. The second unit handles frequently used instructions with 12bit precision very fast, but therefore (-> 12bit!) it can't be used by Pixel Shaders 2.0. Got it? That's why the GeForceFX series (precisely GeForceFX 5200, 5600, 5800) is so slow when we speak about DirectX9 performance! Only the 5700 & 5900 got a 32bit second arithmetic unit, therefore they were able to shorten the performance gap towards ATI.
Moreover, HLSL (High Level Shading Language) has been developed by Microsoft to give game programmers a relativly easy program language for Shaders 2.0. In its first version, the compiler shipping with SDK for DirectX programmers produced code in that order preferred by the Radeon's arithmetic units. Thus, the GeForceFX series got another performance impact, as it prefers another order. There's a new compiler for the SDK now which also can produce GeForceFX preferred code.
I don't want to refer here to ATI's R3XX architecture, let's just say it was more compatible than nVidia's way. Please post here if you want to know more about it, though.
It is now nVidia's job to work closer with game developers, Microsoft & their customers. They optimize their drivers for specific games (e.g. Halo) & for performance. e.g. only the main texture is filtered anisotropic when enabled in some detonator versions. As for ATI, they did the same! And they still do! But nobody cares. Something about "cheating", i was surprised you state that optimizing hardware for games etc. is faint..! That's a stupid opinion man, I want my hardware fast and I don't care if only 3 or 4 of 6 or 8 textures (in multitexturing) are filtered anisotropic 'cause nobody would see any difference!
Listen, I still prefer ATI because the GeForceFX series has some more bugs than the R3XX & their mid-range and budget cards are a lot faster than nVidia's. Not to forget the customer support which is more personal (3dfx-like) than the one of nVidia. And, of course, i hate nVidia for bying 3dfx, like everyone here. But I also know that the GeForceFX series has a great potencial when adressed in the right way, and that's what you just didn't get!

Quote:



Actually, "mines" is the truth...

*sigh*

Of course I know there were meant to be three different products, in fact four were rumored, though a dual Rampage dual SAGE would not have been needed at the time.
And they go like this:

Spectre 1000
Spectre 2000
Spectre 3000

And I know all about the different configs and I know for a fact it wasn't VSA 200...

I'll let you in on a little secret...there is no VSA 200 or even VSA 101 for that fact. It was never called that. But people started calling it so in turn it stuck. VSA100 was a marketing name but Daytona and Rampage had no marketing names yet...

Word of advice, please do not argue with me on this, as this is true...

I think I have an explanation that might clear this whole VSA 200 crap up:

engineering name is not marketing name

the rampage/vsa-200/spectre was still highly in develpment. The enigineers working on the project could have easily called it the VSA-200, and whoever owns the one working card might still call it that. It would be a marketing division of 3dfx that would rename it to something new and killing the voodoo name, but while it was still in development, nobody really cared about that type of stuff, that would be decided later. Anyways, this is just one theory. I have heard all these different...theories, the Spectre series comprised of the rampage/spectre chips, but I've also always heard the rampage also referred to as a VSA-200, even though it would have been a spectre.

Quote:


Must I clarify everything for you?

Quote:


in other words, the rampage chip, aka the VSA 200, which would have been the essential part of the spectre series!

Quote:


sorry, typo on my behalf, I meant Sage!

Quote:


referrring to the fact that the V in VSA stands for Voodoo, but the card would have been a Spectre, not a Voodoo.