OMG nice post Mark! That pretty much sums it up. Nvidia knows how to
on top.
>I think you're mostly right about this, but you characterization of what
>happened with 3dfx is a little too simplistic. They were a small specialty
>company that built add-on PCI 3D-only hardware for *** and built and
>maintained an API to help make this happen. It was apparent that they'd have
>to make a move to AGP-based 2D/3D cards in order to grow beyond this, or
>even to maintain their position - the future of high-end graphics was on the
>AGP bus, which meant incorporating 2D, and other companies were beginning to
>make inroads into 3D from that position. nVidia was the principal
>competition, and the Riva TNT established them as a high-end OEM presence as
>well as a company that could make competent 3D hardware. So 3dfx put out the
>Banshee, which was a middling effort and really just a PCI card using the
>AGP bus, but not a bad start. Then they followed it with the similar but
>faster Voodoo3, a great real-world, bang-for-the-buck card - $130-180 srp
>for V2 SLI-level 3D performance and great 2D. But at the same time they were
>dealing with moving their boardmaking in house, were trying to develop an
>OEM base, Glide was dying, probably erred in putting together their 3500
>flagship card, and weren't focusing enough on their next generation, which
>had to incorporate all those existing features found in competing cards and
>add new ones and even more speed.
>Meanwhile nVidia was putting out the TNT2, which really wasn't anything very
>new either, but it was much more at home in the AGP slot and the press
>jumped all over 3dfx once it became apparent that it could keep pace with
>the V3 in 16-bit benchmarks and could provide playable 32-bit. Then 3dfx
>wasn't doing anything right in their eyes, as I said in another post in this
>thread. The coup de grace was the GeForce, which was rushed out to follow up
>on the TNT2 momentum. The press acted like its hardware T&L was the biggest
>thing since the invention of 3D, and who could tell since there wasn't
>anything to actually run on it? Anyway, the actual hardware trickled out,
>starting with SDR, then DDR, then 64mb of it months later. It almost wasn't
>so much a piece of hardware as a symbol, and the press said if you didn't
>have T&L you might as well box up your PC and start watching TV in six
>months. The opposition didn't really help themselves at that point - the
>Rage Fury Maxx was too slow and flawed, the Savage 2000 sounded good but
>didn't deliver and then S3 was gone, Matrox was silent, and the announced
>Voodoo5 didn't show up at all.
>By the time the V5 did show, nVidia was ready for the GF2, which was the
>real deal. It was faster, it had enhanced features, the drivers had matured,
>and you could actually buy one. The too-little, too-late V4/5 didn't stand
>much of a chance, and it hardly was noticed that T&L games hadn't really
>showed yet. 3dfx was done already, and the 6000 never made it out of the
>lab. But the death knell had rung the generation before, when 3dfx couldn't
>make the transition from a boutique to a mainstream operation. That the GF2
>GTS was very expensive didn't matter much - nVidia had fully established
>their name, so followed the aging TNT2 with the GF2 MX for mid-level buyers
>and the OEMs to spread their base.
>Since then nVidia has backed off their pace. The GF2 Ultra was just a faster
>and more expensive card ($500!) and anyone could have pulled that off, the
>GF3 came out a full year after the GF2 and didn't represent a quantum
>performance leap, and the Ti thing is just a juggling of speeds and prices
>to make it look like they've done something. In comparison, ATi increased
>the speed of their flagship Radeon64 Vivo last year and didn't even mention
>it - with nVidia that would have been a product cycle! The reality is that
>nVidia hasn't made massive progress in the last two years, much less than
>the two years prior to that. The competition hasn't done that well because
>its mostly gone - 3dfx and S3 out of business, and Matrox quietly having
>withdrawn from the high-end sector after dabbling with one card. But ATi has
>come farther in those two years than nVidia, having moved from the Rage
>128-based RF Maxx to the Radeon 8500, every bit the card that the GF3 Ti 500
>is and perhaps more. But they haven't gotten the driver and customer
>relations stuff together the way they should have for this market, and have
>lost a lot of ground to nVidia in the OEM, mobile and Mac sectors.
>So nVidia made their name by defeating 3dfx, a victory largely based on
>flashy features that wowed the press, and on ground that belonged to nVidia
>more than 3dfx. They have parlayed that to a broad base that now goes beyond
>PC video cards. But the image of them as this technological steamroller that
>cranked out new and better cards every six months is a little inaccurate, I
>think.
>> Still, all that being said, it has virtually nothing to do with the
>> reason why 3dfx went out of business. The reason that 3dfx went under
>> is simply that nVidia just blew them away in their core market and
>> 3dfx didn't have a niche to drop back on. S3 also got wiped out and
>> was sold off. Matrox got blown away but was able to fall back on
>> their niche of high-end 2D graphics. ATI was the only company that
>> kept within shooting range, though for a while they were only holding
>> on due to their strong OEM ties and low cost. nVidia simply set a
>> development pace that no one else could match. Everyone else kept
>> planning on building a "TNT killer" or "GeForce killer", but by the
>> time they got to market, nVidia had already moved on by two
>> generations. What's more, nVidia Unified Driver approach gained them
>> some popularity with OEMs and also meant that they were able to
>> quickly develop really good drivers for ALL their operating systems
>> (nVidia to this date remains the only company with really good 3D
>> drivers for Linux, being more then twice as fast as their next
>> competitors on just about everything). ATI was especially bad for
>> this, where their drivers might have worked ok for Win98, but simply
>> sucked for Win2K for quite some time (and under Linux a TNT2 m64 will
>> beat out a Radeon 8500 any day).
>> In short, I don't think that many people can really fault 3dfx that
>> much for their demise, it was really just that nVidia did a LOT of
>> things right and, perhaps more importantly, they did them right now,
>> not two years from now. Even Intel, who many feared would totally
>> dominate the graphics industry in short order was completely blown
>> over by nVidia in the separate graphics chipset market and was
>> relegated to the integrated graphics market (where nVidia is already
>> making a push at as it is).
