All the buzz now are about which card will live to tell a story. In the next months, we will witness the struggle between 3dfx and NVIDIA, each one turning to different sides on the 3D acceleration path. In one side, we have the all-powerful 3dfx with a solid brand name called Voodoo. In the other corner, we can see NVIDIA with a brand-new way of thinking 3D processing. Who will win and why? Who shall deserve our hard-earned money and trust?

3dfx have a long time on this business. We can even say that she’s the real mother of the 3D acceleration on the gaming PC. The Voodoo1 was a board capable of exceeding everything else on it’s time, a real revolution. People all around the world drooled while watching such technology for the first time, and even today, their products are among the best ones, fighting head-to-head for the winner place.

Unfortunately, something seems to be very wrong inside 3dfx by now. Their new board features two technologies that are quite good, but can’t stand the innovation of the competition. To further worsen things, their language concerning the new "breath-taking-achievements" seems to be poised to investors, rather than for developers. While the new texture compression scheme is a real thing, as S3’s Savage showed us the last years, their T-Buffer is a sure shoot in the water, at least the way they are presenting it. 3dfx is claiming that anti-aliasing, motion-blur and depth-of-field are all that it takes to move actual 3D real-time images to the level of pre-rendered animations. Well, every 3D animator in the world knows that’s not true at all. Most animations don’t use motion blur and almost none uses depth-of-field. Anti-aliasing is a sure thing on 3D animations, right, but that doesn’t represent an issue in resolutions above 800x600 on games. It’s dispensable, be sure. And with frame-rates above the 30fps stack, motion blur is not quite a need.

If there is one batch of things capable of turning the actual real-time scenes into pre-rendered ones concerning quality, it’s polygon count, texture quality and raytracing. Raytracing can be left of this article, since the amount of power needed to archiving it is simple ridiculous - we won’t see it in real-time games in the near future, be sure, but a feature of the GeForce - cube environment mapping - can somewhat emulate it. Texture quality points directly to the new compression scheme from 3dfx, and that’s a good thing for them. But in the other hand, high polygon count are a GeForce business, and so, we are left with a duel between texture compression versus polygon count and cube mapping.

Besides that, the GPU scheme of the GeForce enables the computer to be freed of some of the job taken to create 3D scenes, making the system faster as a whole. While this may look and feel like a new concept, it is not. An old computer called Amiga was entirely based in this idea, featuring a separated processor or co-processor for each task - sound, animation, collision detection, graphics and even mouse. And let me say a thing here: it was fast.

The PC is remarkably based on a confusing engineering scheme, rendering some tasks very difficult. Let’s take for example, a Zip-drive operation that renders even the mouse pointer slow or a floppy disk format that makes a 500mhz system slow down as much as 50%: a half-meg memory, 8mhz Amiga could format 4 drives at once while playing digital music, an animation and printing a file - with the mouse still working and asking for more. Try that on a PIII. Furthermore, given the large spectrum of different hardware configurations and brands, the most a 3D board relies on the system, the most it is vulnerable to bugs and failures.

Separate processing is the way to go. The GeForce 256 is the coolest thing to appear in the last years, and if 3dfx are not hiding something really big under their roof as many people believe, things will become dark for them. That’s annoying because people seems to be fighting on 3dfx side: if NVIDIA was about to release a chipset based on a texture compression and blurring scheme while 3dfx presented us with a step towards parallel processing, they would be all screaming that NVIDIA was already dead. Let’s face the truth that the GeForce is brilliant, and will probably open a path to the development of several hardware tricks able to take the PC to a new multitasking level, needed to make better games. This is the point to think about what we really are: brand advocates or good technology advocates.

I take my side on good technology ...