Picture this: 1997. You're walking into your local computer store, the kind with the smell of new plastic and electrical equipment hanging in the air like a specific perfume of the era. Your eyes scan the shelf, and there it is, this sleek black and neon box with the name "Voodoo" printed on the side in letters that seemed to glow with some kind of supernatural promise. The marketing copy on the box promises you "true 3D gaming at last," and you know, somehow, that everything is about to change. You reach for your wallet. This is the moment. This is the card that separates the cool kids from everyone else. This is 3dfx's Voodoo, and for a brief, brilliant moment in the late 1990s, it was the most important piece of hardware in PC gaming.
Today, 3dfx is a cautionary tale, a name whispered in forums by old gamers who remember when the company owned the world. You can't buy a Voodoo card anymore. You can't even run the latest games on one if you still have it in a box in your attic. But the story of how we got here, how a startup founded by people who used to work at Silicon Graphics convinced an entire generation of PC gamers that their lives were incomplete without a dedicated 3D card, is genuinely fascinating. It's a story about vision, about perfect timing, about one catastrophic decision that unraveled an empire, and about the thin line between dominance and oblivion in the tech industry.
The fall of 3dfx is not just about a company that made bad decisions. It's about the complexity of competing in the 3D graphics space, about a market that was moving faster than anyone could predict, and about how a company can be brilliant at one thing and then suddenly irrelevant. But before we get to the collapse, we need to understand how 3dfx got here in the first place.
Where 3dfx Came From
On August 24, 1994, in San Jose, California, three guys from Silicon Graphics decided to start their own company. These weren't random engineers. They knew what they were doing. Ross Tarolli, Gordon Campbell, and Gary Smith had worked at SGI, where they'd spent years thinking about 3D graphics and how to make them fast. They understood, at a level that most people didn't, that PC gaming was about to explode. The home PC market was finally getting powerful enough that serious 3D graphics were possible, but nobody had really figured out how to do it efficiently yet. The three founders looked at the market and saw an opportunity that nobody else seemed to notice.
The company was called 3dfx Interactive, and the name itself was meant to evoke something futuristic and powerful. But the really important thing about those early days wasn't the name or the founding story, though that stuff matters. What mattered was that 3dfx attracted the right kind of investment. Gordon Campbell at TechFarm believed in these guys enough to put real money behind them. This wasn't some garage startup that got lucky. This was smart people with real experience backed by investors who understood the industry. From day one, 3dfx had credibility and resources.
In those early years, 1994 and 1995, the PC graphics landscape was a complete mess. You could buy a Matrox card, or a Tseng card, or a Trident card, and none of them could really do 3D the way you needed for modern games. Games were still mostly running on the CPU, doing the 3D calculations in software, which meant everything was slow and limited. What 3dfx understood was that the future belonged to companies that could offload the 3D work onto specialized hardware. They weren't the only company thinking this way, but they were the ones who executed best.
The company spent its first couple years in stealth mode, basically, developing the technology that would change everything. They weren't making announcements or trying to drum up press coverage. They were heads-down on engineering, trying to solve the problem of how to make 3D graphics fast enough for real-time gaming. And by 1996, they had something.
The Voodoo Changed Everything
When the Voodoo1 finally shipped in October 1996, it wasn't available as a standalone product you could just walk into a store and buy. It came integrated into a card made by Orchid, a company that was one of the original PC graphics card makers. The card was called the Orchid Righteous 3D, and it retailed for around $299. That was a lot of money in 1996, roughly equivalent to $600 in today's money. This was a serious purchase. You didn't buy a Voodoo card on a whim.
But the thing about the Voodoo was that it didn't replace your existing graphics card. This is important to understand, because it was strange and unintuitive. Your existing card still handled your 2D graphics, your Windows desktop, your screen resolution and refresh rate. The Voodoo card was connected via a passthrough, which meant it sat between your main card and your monitor, intercepting the 3D data but letting the 2D stuff pass through. This was weird engineering, and it meant you needed to have the right compatible card, and you needed to deal with some cables and configuration, but it worked. And more importantly, it meant that game developers finally had hardware that could actually deliver 3D graphics at speed.
The real magic, though, wasn't just the hardware. It was the software interface that 3dfx created to go with it. They called it Glide, and Glide was revolutionary. See, at the time, there was this graphics standard called OpenGL that everyone was supposed to use, and it was pretty comprehensive but also pretty complex. 3dfx looked at OpenGL and decided that what the gaming industry really needed was something simpler, something that was specifically designed for fast, streamlined 3D graphics without all the bloat. So they built Glide. And they gave it away to developers for free.
The first game that really showed what the Voodoo could do was a modified version of id Software's Quake. John Carmack, the legendary programmer at id, worked with 3dfx to create a version of Quake that used Glide instead of the existing 3D renderers. This wasn't just a port. This was Quake looking better, running faster, and basically showing the world what PC 3D gaming could be if you had the right hardware and the right software working together. People saw GLQuake running on a Voodoo and they basically lost their minds. This is what they'd been waiting for. This is what they'd been dreaming about.
The response was extraordinary. Within months, other games started getting Glide support. Descent, Descent 2, Jedi Knight, Quake 2, all the major games of the era started shipping with Voodoo support. It became this thing where if you wanted to play the latest games with the best visuals and the best performance, you needed a Voodoo card. The gaming community coalesced around 3dfx not because they had some regulatory advantage or because they had some evil monopoly practices, but because they had the best product and they'd made it easy for developers to support it. This is the part of the story where 3dfx did everything right.
Total Market Domination
By the end of 1997, 3dfx controlled somewhere between 80 and 85 percent of the dedicated 3D graphics card market. This is not a small number. This is domination. This is the kind of market share that makes everybody else in the industry nervous and makes your company feel invincible. And in 1997, 3dfx went public. June 25, 1997, to be exact. The IPO was priced at $11 per share, and it raised about $33 million for the company. The stock market was telling 3dfx that this was the future, that they were sitting on something special, that this company was going to be huge.
And here's where it gets interesting: the revenue numbers back that up. In 1996, 3dfx had revenues of about $44.1 million. By 1997, that number had jumped to $202.6 million. We're talking about more than quadrupling revenue in a year. This is the kind of growth that venture capitalists dream about. This is the kind of growth that makes investors believe they've found the next thing. The company was riding a wave, and nobody seemed to think the wave would ever crash. Every month, more people bought Voodoo cards. Every month, more games got Glide support. Every month, 3dfx looked more and more like the company that was going to define the future of gaming.
The stock price reflected this euphoria. People who bought the IPO and held on saw their investment multiply many times over. There were stories in tech magazines about how 3dfx was going to be the next Intel, the next Microsoft, the next huge tech company that would be around for decades. The narrative was almost written. The only thing that could stop 3dfx was if they made some kind of serious strategic mistake, if they got distracted, if they took their eye off the ball. And spoiler alert: that's exactly what happened, but we're getting ahead of ourselves.
In 1997, 3dfx was the undisputed king of PC 3D graphics. They had enemies, certainly. Matrox was still making cards. NVIDIA was working on something. ATI was in the market. But none of these companies could touch 3dfx's market share. None of them had the same developer support. None of them had convinced gamers the way 3dfx had that their card was the essential purchase. When you walked into a computer store in 1997 and looked at the 3D graphics cards, the Voodoo was what everyone wanted. The passthrough arrangement was a little weird, sure, but the performance was so much better than everything else that the weird arrangement didn't matter. You put up with it because the payoff was worth it.
The Voodoo2 and Peak Power
In March 1998, about nine months after the IPO, 3dfx released the Voodoo2. If the original Voodoo had been a revelation, the Voodoo2 was absolutely dominant. The card was faster, more capable, and it introduced something called SLI technology, which stood for Scan Line Interleaving. This meant you could put two Voodoo2 cards in your system at the same time, and they would work together to render a single image. This was wild. This was sci-fi territory for consumer hardware. You could basically double your 3D performance by buying two cards instead of one. It was insane, but it worked, and it was thrilling to gamers who had the money and the ambition to do it.
The Voodoo2 also brought 1024x768 resolution support, which meant you could finally play your 3D games at a resolution that didn't look like it had been rendered for a monitor from 1985. The original Voodoo had been limited to 640x480 or 800x600, which looked okay if you squinted but wasn't exactly gorgeous. The Voodoo2 changed that calculus. You could play Quake 2 or Unreal at 1024x768 with Voodoo2 SLI and it would look incredible. It would look like the future was here. And for a brief moment, it actually felt like that's where the technology was heading.
This is also when the LAN party culture really exploded. If you were a gamer in 1998, 1999, you probably went to at least one LAN party, where gamers would haul their computers to some convention center or university and all plug into a local network to play games together. Quake 2 was the game, and if you showed up to a LAN party with a Voodoo2 SLI system, you were basically a god. You had the hardware that everyone wanted. The Voodoo2 became synonymous with competitive gaming, with serious gamers, with people who took PC gaming seriously. It wasn't just a graphics card. It was a status symbol. It was proof that you cared enough about gaming to invest serious money into your rig.
The Voodoo2 sold phenomenally well. 3dfx's market share remained in that 80-85 percent range. Revenue continued to climb. The stock price continued to climb. Everything looked good. Everything looked like 3dfx was going to keep dominating forever. But here's the thing: they were already making the decision that would unravel everything.
The STB Mistake That Started the Collapse
On December 14, 1998, 3dfx announced that they were acquiring STB Systems for $141 million. STB was a company that made graphics cards, but they also had manufacturing capability, they had relationships with PC manufacturers, they had a distribution network. On the surface, the acquisition made sense. 3dfx wanted to move beyond being a chipmaker to actually being a graphics card manufacturer. They wanted to own the entire supply chain. They wanted to be more integrated, more in control of their own destiny.
This was the moment where everything started to go wrong. Here's why: the OEM partners that 3dfx had been working with, companies like Diamond Multimedia that made the Diamond Monster line of Voodoo cards, they were already uncomfortable. They made their living selling 3dfx chips as finished graphics cards. The more that 3dfx integrated vertically, the more they became a competitor to these OEM partners rather than just a supplier. Diamond Multimedia had been one of the biggest Voodoo enthusiasts, making some of the most popular versions of the card. But after the STB acquisition, Diamond saw the writing on the wall. 3dfx was going to compete with them directly. So Diamond started looking for alternatives. They started working with NVIDIA.
And here's where it gets worse: the STB acquisition was expensive, and it pulled 3dfx's focus away from what they should have been doing, which was staying ahead in the 3D graphics arms race. The company suddenly had to manage a manufacturing operation, deal with supply chain issues, handle inventory management. These are not things that a graphics chip company should be spending its time on. These are distractions. And while 3dfx was dealing with the STB integration, the competitive landscape was shifting.
NVIDIA, a company that had been working in relative obscurity up until this point, was preparing to release something called the GeForce 256. NVIDIA had been founded in 1993, so they were only a few years younger than 3dfx, but they'd been keeping their head down, developing technology, building relationships with hardware manufacturers. They didn't have 3dfx's market dominance, but they had something that was about to become more important than market dominance: they had support for Direct3D.
NVIDIA and the End
Direct3D was Microsoft's graphics API, and it was the counterweight to Glide. While 3dfx had built Glide as this streamlined, simple interface for 3D graphics, Direct3D was more comprehensive, more heavyweight, but more importantly, it was backed by Microsoft. And Microsoft had this thing called Windows, which was running on pretty much every PC in the world. As the late 1990s progressed, the idea of using a proprietary API like Glide seemed increasingly risky. What if Microsoft decided to do something that hurt Glide? What if Direct3D got better tools and better support? What if game developers started seeing Glide as a limiting factor rather than a feature?
The GeForce 256, released in 1999, supported Direct3D. It wasn't dramatically faster than a Voodoo2, but it was competitive, it was available from multiple manufacturers who weren't owned by NVIDIA, and it had the backing of Microsoft. And here's the thing that really mattered: it didn't need a passthrough cable. You plugged it into your PCI slot and it handled both 2D and 3D graphics. It was simpler. It was cleaner. It was more integrated into the Windows experience.
Meanwhile, 3dfx was working on the Voodoo3 and the Voodoo5, which were supposed to combine 3D and 2D graphics on a single card, finally getting rid of the passthrough arrangement. But these cards were delayed. The engineering was harder than expected. The market was moving faster than anyone had predicted. And when the Voodoo3 finally came out, it was too late. NVIDIA had momentum. NVIDIA had developer support that was growing by the day. NVIDIA had a clear strategy. 3dfx, meanwhile, was still dealing with the STB integration, still trying to figure out how to manufacture cards efficiently, still dealing with all the chaos that comes from a merger.
And here's where it gets really interesting: 3dfx's core problem wasn't that they made bad products. The Voodoo3 and Voodoo5 were solid cards. The problem was that the market had shifted underneath them. The idea that you needed a specialized 3D graphics card was becoming less necessary as integrated graphics got better and CPUs got faster. The idea that you needed a proprietary API like Glide was becoming quaint as Direct3D got better and OpenGL continued to improve. 3dfx had built their dominance on these two things: specialized hardware and a proprietary API. But the industry was moving toward open standards and integrated solutions. And 3dfx couldn't pivot fast enough to keep up.
By 2000, it was clear that 3dfx was in serious trouble. The stock price had crashed. The revenue had stopped growing. The market share was eroding. Diamond Multimedia had completely abandoned them. Other OEM partners were following suit. And NVIDIA, this company that nobody had really heard of a few years before, was becoming the dominant player in 3D graphics. It happened in what felt like an instant, which is kind of insane when you think about it. You go from owning 80 percent of the market to being in a death spiral in less than three years.
In the fall of 2000, NVIDIA approached 3dfx about an acquisition. The deal was for around $112 million, which sounds like a lot until you remember that 3dfx had raised $33 million in their IPO and had been making hundreds of millions of dollars in revenue. The company that once seemed like it was going to be the next Microsoft was being bought for a fraction of what the market had once valued it at. Shareholders who had bought the IPO watched their investment crater. Employees who had held company stock watched their stock options become worthless. The gold rush was over.
And here's the thing that really stung: NVIDIA didn't even want to keep 3dfx operating as a company. They wanted the technology, they wanted the patents, they wanted to integrate some of the engineering talent, but they didn't want to be in the business of manufacturing graphics cards based on 3dfx's architecture. They wanted to kill Glide, they wanted to move everyone over to Direct3D and OpenGL, they wanted to make it clear that the 3dfx era was over. On February 15, 2001, NVIDIA officially ceased support for the Voodoo line. Any Voodoo card sitting in a system on that date suddenly became a historical artifact rather than a usable piece of hardware. The Glide API had been supplanted. The era had ended.
And exacerbating all of this was the formal bankruptcy filing. On October 15, 2002, 3dfx formally filed for bankruptcy. This was years after the NVIDIA acquisition, years after the company had effectively ceased to exist as an independent entity. But the bankruptcy filing was the official acknowledgment that the company was gone. All the promises of the 1997 IPO, all the hype about being the future of gaming, all the dominance of the late 1990s, it all came down to this: a bankruptcy filing, a company that no longer existed, a footnote in technology history.
What 3dfx Left Behind
The collapse of 3dfx is taught in business schools as a case study in how to squander market dominance. How do you go from owning 85 percent of a market to bankruptcy in four years? How do you make that many strategic mistakes? And the answer, of course, is complicated. The STB acquisition was the obvious mistake, the moment where the company took its eye off the ball. But there were other factors too. The company was growing too fast. The executive team might not have had the right expertise. The board might not have been asking the right questions. And the market was moving faster than anyone had predicted. Direct3D and OpenGL were improving faster than anyone expected. NVIDIA was executing better than anyone expected. Sometimes, the thing that makes you the dominant player in one era makes you obsolete in the next era. The very specialization that had been 3dfx's strength became their weakness.
But the interesting thing about 3dfx's legacy is that they didn't go to waste. NVIDIA bought their patents, bought their technology, bought some of their people. And if you trace the lineage of NVIDIA's technology forward, you can see the fingerprints of 3dfx all over it. The things that 3dfx learned about efficiently rendering 3D graphics, about building specialized hardware, about serving the gaming market, those insights didn't disappear. They got absorbed into NVIDIA. And NVIDIA built on those insights, improved on them, and became the dominant player that 3dfx had been.
And there's something else too: the nostalgia market. Today, there are retro computer enthusiasts who collect Voodoo cards, who get them working in old systems, who try to play the games from the 1990s the way they were meant to be played. These cards are valuable now in a way that has nothing to do with their technical capabilities and everything to do with what they represent. They represent a moment in time, a pivotal moment when PC gaming changed forever. They represent the late 1990s, the LAN parties, the excitement of seeing Quake 2 running at 1024x768 with real-time 3D for the first time. If you want to experience what gaming was like in 1998, you need a Voodoo2 and a Voodoo2. There's no digital equivalent. You can't just run these games on modern hardware and get the same experience. The technology was so specific to that era that it's become historical.
The fall of 3dfx also serves as a reminder of how quickly things change in the technology industry. There are no permanent winners. There's no guaranteed success, no matter how dominant you are. 3dfx had the best product, the most developer support, the highest market share, and a huge IPO. And in less than five years, they were bankrupt. It's a humbling thing to think about if you're running a tech company. It's a reminder to stay focused, to keep innovating, to not get distracted by things that aren't core to your mission, to pay attention to what your competitors are doing. 3dfx failed at most of these things. And the industry moved on.
FAQ
Q: Can you still buy a Voodoo graphics card?
A: Not from any manufacturer. 3dfx is gone, and NVIDIA stopped production of Voodoo cards years ago. However, you can find used Voodoo cards on the secondhand market, eBay, and retro computer forums. Prices vary wildly depending on the model and condition. A Voodoo2 in good condition can fetch hundreds of dollars from collectors.
Q: What's the difference between the Voodoo1, Voodoo2, and Voodoo3?
A: The Voodoo1 was the original, released in 1996 via the Orchid Righteous 3D. It required a passthrough cable to work with your existing graphics card. The Voodoo2, released in 1998, was faster and introduced SLI technology and 1024x768 resolution support. The Voodoo3, released in 1999, was supposed to combine 3D and 2D graphics on a single card and eliminate the passthrough arrangement, but it came too late to save the company.
Q: Did NVIDIA use 3dfx's technology?
A: NVIDIA acquired 3dfx and its patents in 2000. They integrated some of the technology and engineering talent into NVIDIA's own graphics architecture. However, Direct3D and OpenGL became the dominant APIs rather than Glide, so the proprietary parts of 3dfx's technology weren't as important as the patent portfolio and engineering expertise.
Q: Why did Glide fail?
A: Glide was a proprietary API created by 3dfx. While it was streamlined for gaming, the industry ultimately embraced open standards like Direct3D and OpenGL. As these standards improved, the incentive to develop specifically for Glide decreased. Microsoft's backing of Direct3D, combined with its dominance in the PC operating system market, meant that supporting Direct3D was more practical for game developers than maintaining Glide support.
Q: Are there any games that still require a Voodoo card?
A: Not in the modern sense. Games from the late 1990s were optimized for Voodoo hardware, but they can all run on modern systems through emulation or through ports to modern APIs like OpenGL. If you want to play these games on a Voodoo card the way they were originally intended, you would need an old PC with the appropriate hardware and operating system.
Q: What was the STB acquisition?
A: In December 1998, 3dfx acquired STB Systems, a graphics card manufacturer, for $141 million. This was intended to help 3dfx move from being primarily a chipmaker to being an end-to-end graphics card manufacturer. However, the acquisition distracted the company from its core business, and it alienated OEM partners like Diamond Multimedia, who became competitors instead of allies. Many analysts consider this acquisition to be the critical mistake that doomed 3dfx.
Q: When did 3dfx go bankrupt?
A: 3dfx filed for bankruptcy on October 15, 2002. However, the company had effectively ceased to exist much earlier. NVIDIA acquired 3dfx in 2000, and support for Voodoo cards was discontinued on February 15, 2001. The bankruptcy filing was more of a formal legal acknowledgment of what had already happened.