Isn't it interesting how a company can have a near-monopoly on the 3D accelerator market and then suddenly go bust? 3DFX was a prime example of how NOT to manage a company, or to solely rely on marketing hype to sell video cards. 3DFX committed a number of blunders; while there was no single cause of the fall of 3dfx, the following certainly didn't help
- They stopped outsourcing their video cards. After 3dfx's acquisition of STB, they stopped letting other manufacturers produce their video cards; the only way to get a Voodoo card was through STB or direct from 3dfx. Needless to say, this did not have a good impact on their market share, as prices rose shortly after this decision
- They made excuses to justify their underpowered hardware. When the Voodoo3 came out, it was a major disappointment. While the competition had support for texture sizes above 256x256, and had 32-bit colour, 3dfx's Voodoo3 was simply a souped-up version of their integrated 2D/3D Banshee, which in turn was simply an improvement on the original Voodoo, which was released well over three years ago. 3dfx justified this by saying that gamers shouldn't be concerned with eye candy, 32-bit was overrated since there were no games supporting it, etc. They also refused to support AGP texturing in any way, shape, or form, as evidenced by AnandTech's 3dfx Interview from September 1999
Will we finally see a 3dfx chip with full AGP texturing support?
We will support full AGP texturing in forthcoming products. However, it really is a moot point at this stage because AGP texturing is not being embraced (nor has it ever been) by the development community. There simply is not enough bandwidth on the AGP bus, even with AGP 4x, to sustain high fill-rates when texturing from AGP system memory. What has happened is that as polygonal complexity increases, the additional AGP bus bandwidth is used up for polygon traffic. And, the continual decline in memory prices has made having to store textures in system memory less and less of an issue (witness that you can buy 32MB graphics boards now for under $100). And, even Intel is backing off from its strong support of AGP texturing for the same reasons I outline, so I think you’ll see moving forward it becoming a very unimportant "checklist" feature
- They couldn't keep up with the competition. 3dfx even went so far as to include apology notes in their Voodoo5 boxes apologizing for the continually delayed (and canned) Voodoo 5 6000
- The Voodoo5 was a disaster. Had the Voodoo5 been released on time, it would have crushed whatever competition there was in terms of raw power. However, it was delayed so much that by the time it was finally out the door, it was competing against the GeForce2 line of cards. Consequently, it got squashed in the benchmarks. It really says something when a top-of-the-line card is barely beaten by the GeForce2 MX value card in benchmarks.
However, that's not the end of it. The Voodoo5 5500 required a power connecter direct from the power supply, which was a first in video card history. 3dfx justified this by saying that as cards got faster and faster, more power would be needed to sate the card's demands. Of course, this claim turned out to be utter bullshit since the GeForce4 Ti 4400 gets along just fine with the AGP Pro specification
T-Buffer was supposed to revolutionize 3D rendering - except it didn't. A small program named 'T-Bluffer' was written to show what lies 3dfx were spewing forth; it did a good job of simulating 3dfx's widely-touted 'exclusive' T-Buffer. Indeed, 3dfx's strategy seemed to be simply to ignore the feature set and throw more VSA-100 chips on the board. After all, it's supposed to be scalable.. and true, VSA was scalable. The problem was that it didn't scale well, as the VSA-100 chips required a fan on them since they generated stupendous amounts of heat
- Botched OpenGL support Even today, developers must continue to support the 3dfx legacy through the cursed 3dfxgl.dll driver, reflecting 3dfx's tacking on OpenGL support at the last minute. This happened after Carmack refused to optimize Quake for Glide, which turned out to be a wise decision