Originally Published: Wednesday, 17 January 2001 Author: Ross Sanders
Published to: enhance_articles_hardware/Hardware Articles Page: 1/1 - [Printable]

A Requiem for 3dfx

It's time to wave goodbye to the venerable 3d graphics innovator called 3dfx, according to this hardware article written by Ross Sanders. Last month the company announced plans to sell its assets to nVidia, bringing to an end the spirited rivalry between the two. What does it mean for customers?

   Page 1 of 1  

Say goodbye to long time 3d graphics hardware innovator 3dfx. For those not keeping up with PC news, 3dfx announced Dec. 15, that it is selling off its assets to long-time rival nVidia. This move, not wholly unexpected from a financial standpoint, comes as somewhat of let down for 3dfx fans, including me.

There's no question that 3dfx has been technically one-upped for the last few years by its many competitors. In 1998 the Voodoo 3 was released and was quickly overcome in performance and video memory by nVidia with the TNT2. The cycle once again continued with the outrageously expensive Voodoo 5 last year, one upped by nVidia's GeForce2.

One of the things, other than performance, that apparently sunk 3dfx's boat was the insistence on maintaining the Glide standard for its own hardware but tying up licensing to competitors to the point where no one wanted to deal with them. 3dfx obviously was trying for a market monopoly on the 3d accelerator front by keeping Glide, which was a very good API, tied to its chipsets. For a while it actually did have more or less a monopoly since its only competing API was Direct3D from Microsoft. Most games that included 3d accelerator support only supported the Glide API. That meant you had to have a Voodoo chipset in your computer.

Other chipset manufacturers, in order to break 3dfx's monopoly on 3d gaming knew that Direct3D could never compete with Glide. Instead they turned to SGI and its OpenGL standard that had long been known in the Unix world as the standard for 3d API. Enter OpenGL drivers for Windows and the death knell for Glide and 3dfx. Since 3dfx refused to license its own technology and software to its competitors, SGI was more than happy to fill in the serious gap. By last year, games began shipping with the OpenGL API to the exclusion of Glide. 3dfx's own OpenGL drivers were naturally different than anyone else's so game developers had once again two APIs to develop. On top of that the Voodoo chipset, performance-wise, was no longer top dog.

Even though 3dfx made a good run of the game of hardware production, they failed to heed the emergence of new standards. They missed the opportunity to license their own technology in a way that other companies could take advantage of it and increase 3dfx's coffers in that way when sales of the Voodoo-based boards declined.

Standards are standards. Companies either create them or conform to them. Those that try to go their own separate way generally end up doing poorly. The lesson to be learned from 3dfx's failure is to allow open standards. That is, open in the sense that others can use them. An open standard does not have to be free of charge. OpenGL is an open standard that is hardware independent but is not free of charge. Good technology is like money: it does no one a bit of good by sitting on it. The only way to see a return on money is to invest it wisely. If the technology, such as Glide and OpenGL, is good enough, companies will be willing to pay to use it. If it is not or it's parent developer is stingy, it will fall by the wayside, at the most a footnote in computer history.

What else can we learn from 3dfx? People like standards. Amid the chaos of computer technology and operating systems with a list of names that can reach to the moon, people like finding similarities that they can build on. Linux needs more cohesiveness between it's disparate distributions before it will seriously take on the desktop industry. Linux needs standards. As it presently stands each distribution basically does it's own thing. Each has it's own way of laying out files. Each has it's own way of placing startup scripts. The list is endless. The only thing that is remotely standard between them is the Linux kernel itself. This is one reason that companies often do not wish to publish works for the Linux market. Support is a nightmare of package managers, file layouts, strange kernel patches, and weird rc layouts.

A Linux standards committee would be the perfect tool to overcome Tux's chaotic beginnings and mold our tribes into some resemblance of order. This committee could outline a basic layout for a binary distribution with things that could make life easier for all concerned. Those distributions in compliance with the easy guidelines would receive a stamp of approval from the committee. Those that do not, let the buyer/user beware. These standards would not impinge on the creativity that has made Linux what it is, but simply make things easier for those who wish to write software to do so without having to write not only the source code, but 3 different package managers, innumerable sub-packages for Debian, Storm, Red Hat, Mandrake, SUSE, and the list goes on and on and on.....

Ross will be missing 3dfx, despite his criticism. The king is dead, long live the king! I wish nVidia well in its current and future endevours. wolf_fire@linux.com





   Page 1 of 1