Every time I sit down in a movie theater and someone hands me a pair of those cheap plastic 3D glasses, I smile. Most people think modern 3D movies started with Avatar. I know better. I was there when the whole thing was three beige computers, a curved screen, and a golden death mask floating in the middle of a room.
Let me back up.
What real 3D looked like in 1998
In the late 1990s, real 3D, the kind that actually worked and not the red-and-blue cardboard stuff, lived in exactly one place: industrial visualization labs running Silicon Graphics supercomputers. An SGI Onyx2 was a refrigerator-sized machine that cost millions, and the software that drove it came from a handful of companies. The biggest one in Europe was Prosolvia, out of Gothenburg, Sweden. At its peak Prosolvia was worth over 4 billion Swedish kronor, roughly 500 million US dollars at the exchange rates of the day, and if you were doing serious VR simulation on that side of the Atlantic, you were probably using one of their products.
CrystalEyes: the glasses before the glasses
The glasses we wore back then were nothing like what you get at the theater today. They were called CrystalEyes, made by a company named StereoGraphics, founded by Lenny Lipton. Remember that name, he comes back at the end of this story. CrystalEyes were heavy, beige, battery-powered active shutter glasses. Each lens was a liquid crystal panel that could snap from clear to black. The monitor blasted alternating left-eye and right-eye images at 120 times a second, an infrared emitter sitting on top of the monitor flashed timing pulses across the room, and the glasses shuttered each eye in perfect sync with the screen. When it worked, it was magic. When you turned your head away from the emitter and broke the sync, the whole world dissolved into harsh flicker. That flicker was the tax you paid for seeing the future.
King Tut’s tomb, scanned a decade early
The showpiece I remember best was King Tut. In May of 1998, under Prosolvia, we provided the engine for an interactive walkthrough of the tomb of Tutankhamun. It was the crown jewel of the Mysteries of Egypt exhibition at the Canadian Museum of Civilization in Hull, Quebec.
Understand what that meant in 1998. Nobody was laser mapping world heritage sites yet. The famous documented scanning projects inside that tomb didn’t start until roughly a decade later. The National Research Council of Canada had to build their own custom laser triangulation scanners. Their Visual Information Technology group took those scanners to Egypt and mapped the structural geometry of the burial chamber into millions of points of data, then wrapped that geometry in some of the highest resolution photography anyone had been allowed to shoot in the tomb.
Here’s the part most people miss. They didn’t just stitch photographs together and call it 3D. They captured the spatial depth first. Geometry, then texture. That distinction is the whole reason the thing read as a place instead of a picture.
This was genuinely first, and almost nobody wrote it down, because that’s how it went back then. Companies built jaw-dropping demos to win contracts, showed them to a room full of clients, and never published a word.
December 1998: four billion kronor, gone
Then, in December 1998, Prosolvia collapsed. Sudden bankruptcy, accusations flying about how revenue had been booked. A 4-billion-kronor company, gone almost overnight. For the record: the Swedish courts fully cleared the founders in 2005 and ruled the contested deals legitimate, but by then the original company had been dead for seven years. That’s the part of this business nobody warns you about. The technology can be perfect and the company can still die.
Irvine, California, and a browser 3D engine in 1999
Here’s where it gets good. The founders bought the source code back from the Swedish bankruptcy trustees, packed up, and landed in Irvine, California. Two ventures launched at once and I was pulled into what would become the dot-com bubble.
On top of forming EON Reality, they ran a short-lived venture called RealityBUY.com, whose idea was streaming interactive 3D product models into a web browser for online shopping. In 1999. That’s WebGL, more than a decade before WebGL existed. It actually worked. They produced plug-ins for Netscape and Internet Explorer, then made a deal with Macromedia to have the engine included in the Flash distribution, so if you had Flash it would load the EON 3D engine and let you turn a product around on the page. They even had a major client in Office Depot, who integrated it into their furniture site. After the bubble burst, what was left was EON Reality and a Windows-based 3D simulation engine.
The Concave: three PCs against a supercomputer
And in 1999, EON built the thing I still think about: the Concave.
The prevailing wisdom said you needed that multi-million dollar SGI machine to do immersive 3D. The Concave said no you don’t. Three ordinary Windows NT PCs. Three NVIDIA workstation cards at about $1,200 apiece, the very first generation of Quadro cards. A custom cable physically linking the three cards together with special driver code from NVIDIA, so all three machines swapped their frame buffers within microseconds of each other. Three high-end Barco projectors throwing onto a 120-degree curved screen with a 10-degree overlap between them, and all the geometric warping needed to bend three flat images onto a curve handled inside the Barco hardware itself, so the PCs didn’t waste a single cycle on it. And synced on top of all of that: CrystalEyes shutter glasses, genlocked across the whole rig, so the stereo held flicker-free across a curved wall of light.
What it cost, and why the number is the point
Total cost including the screen, the projectors, the software and the glasses was just over what you would have paid for a single well-equipped SGI Onyx2 alone. About $750,000.
Read that again, because the number is the entire point and it is not the point people expect. The Concave was not cheap. It cost about what one SGI cost. What changed was what that money now bought you, and what it was built out of. Parts anyone could order. In price alone this put museum-grade and industry-leading 3D simulation within reach of buyers who were never going to sign off on a supercomputer. That one demo room in Irvine was the future of the entire industry arriving early: commodity PCs killing the UNIX supercomputer. Within a few years SGI was a memory, and NVIDIA, which picked up a large block of SGI’s graphics patents on the way through, is now one of the most valuable companies on earth.
The mask that floated
The demo we ran on it? King Tut. The old Prosolvia dataset, ported off the SGI and onto Windows, polygon budgets squeezed until it fit. A narrated tour through the tomb, ending on a close-up of the golden death mask.
And because the curved screen filled your peripheral vision, with no monitor bezel, no edge, no frame of reference telling your brain “this is a screen”, when we pushed the stereo coordinates into negative parallax the mask came off the screen. It detached from the background and hung there, unsupported, in the middle of the room. Then it started to rotate. Grown adults reached out to touch it. Every single time.
That’s what being on the bleeding edge felt like. Not a press release. A room full of people reaching for something that wasn’t there.
The line from that room to your local cinema
Now the ending, and why I smile at the movie theater. In 2005 a company called RealD bought StereoGraphics, the CrystalEyes people, for the patents and the engineers. Lenny Lipton, the man who invented the glasses I wore in that demo room, became RealD’s Chief Technology Officer. They took a StereoGraphics device called the ZScreen, an electro-optical modulator originally built to sit in front of a workstation monitor, and scaled it up to sit in front of a digital cinema projector, flipping polarization at 144 frames per second. Which meant theaters could finally ditch expensive shutter glasses and hand out featherweight polarized plastic ones instead. Chicken Little tested it in 2005. Avatar detonated it in 2009. Today it’s just going to the movies.
So the line runs straight and unbroken: the beige glasses on my face in that demo room, to the modulator on the monitor, to the modulator on the projector, to the glasses in your hand at the multiplex. Same company, same inventor, same physics, scaled up to the whole world.
Why almost none of this is written down
Here’s the thing about the bleeding edge: the world doesn’t keep records of it. Try to find the Prosolvia King Tut demo online today. You can’t. Not in the trade press, not in the academic papers, not in the web archives. The company died before the internet learned to remember things, and demo rooms don’t publish proceedings. The public record of some of this work is people like me, who stood in the room and watched a dead pharaoh’s golden face float in the air, years before the rest of the world put on the glasses.
I got to be there first. That never stops being cool.
If you were there
I was one of the people who built this, so treat me as a primary source with all the problems that implies. I have checked what I could check and flagged what I couldn’t. If you worked at Prosolvia, EON Reality, SGI or StereoGraphics in this era and I got something wrong, say so in the comments on the video and I will correct it here. I am trying to get this on the record properly, not just tell a good story.
Rob streams live repairs and builds every Tuesday and Saturday at 7pm Eastern on Twitch. His current series is Shelf Life, pulling broken hardware off a ten-year shelf and not stopping until he knows why it died. Premieres September 12, 2026.