Prepare to Meet Thy Doom

Written by David Kushner, Wired

Thursday May 1st, 2003

“How are the fingers?” coder Jim Dosé asks artist Kenneth Scott, as they stand in the kitchen of id Software’s Mesquite, Texas, headquarters.

“Shattered,” Scott replies wearily, waving a splint – the result of a rare office football game played to ease tension. But he’ll type with the eight digits that work. Lead designer Tim Willits hobbles in with a thigh of busted capillaries from the same game. The art guys just scanned his wound to use as skin for a monster.

These days, the employees at id need to play with pain. They’re hard at work on Doom III, which is already a shoo-in for event of the year in the $10.8 billion videogame industry, even though it’s not expected out until fall. In the dozens of times I’ve come here to research Masters of Doom, my book about John Carmack and his ex-partner, John Romero, I’ve never seen id’s office as focused as it is now. There are no CDs whizzing into walls like Frisbees. No keyboards being hammered during Quake III marathons. No screams of the traditional shooter deathmatch taunt, “Suck it down!” Right now, the only things being sucked down are the brownies and coffee in the kitchen – a caffeine-sugar slammer to fortify the troops for yet another late night.

A $108 million brand (counting the first two titles and various expansion packs), Doom napalmed the path for everything that followed: the first-person shooter action of Halo, the Internet play of EverQuest, the ultraviolence of Grand Theft Auto III. Doom was the first product to invite gamers to get under the hood and fiddle around with accessible, adaptable code that allowed for modifications, or mods, and there are versions based on everything from Star Wars to Aliens. As Doom and its successors became gaming standards, companies like Valve and Raven licensed id’s graphics engines to create their own shooters.

Like the original, Doom III casts the player as a lone space marine in a Martian military base where a group of scientists have unwittingly opened a portal to hell. But it’s hardly a rehash. The new game sets a benchmark for computer graphics, just as so many of Carmack’s innovations have done in the decade since id released the first Doom. By ushering in the age of 3-D acceleration with Quake in 1996, Carmack all but invented the graphics card industry currently dominated by Nvidia. Today, at a time when few applications stretch the capabilities of a $500 PC, Carmack creates programs that require high-end systems. “Carmack’s games always push the envelope,” says David Kirk, chief scientist at Nvidia. “Doom III is the harbinger of technology to come.”

These days, the game is also the harbinger of all-nighters. During slow times, Carmack, who builds high-powered rockets with pet project Armadillo Aerospace in his spare time, leaves at 2 am (he usually arrives around noon). The hard work is paying off. A mere demo was Best of Show at last year’s Electronic Entertainment Expo. Rumors abound that this could be Carmack’s last major work. Doom III, says Matt Helgeson, senior editor of Game Informer, “has slightly lower expectations around it than the second coming of Jesus.”

The brownies vanish. The coffee drains. The pressure resumes. As level designer Christian Antkow says, “We cannot fuck this up.”

From the day his mother took him for a TRS-80 programming course when he was in the fifth grade, John Carmack dedicated himself to creating compelling computer graphics. After being thrown into a juvenile home for stealing an Apple II at age 14, he took the opportunity to create Wraith, a sprawling role-playing game that emulated the ambitious Ultima franchise. In 1990, while working at Softdisk, the Shreveport, Louisiana-based software company where he met his future id partners, the 19-year-old Carmack figured out how to bring side-scrolling to the PC so he could re-create the arcade action of Super Mario Brothers 3. He used the same breakthrough on id’s first best-seller, Commander Keen.

Since then, Carmack has written a new graphics engine for almost every product he’s developed. In an essentially visual medium, the graphics engine – the core code that determines how images are displayed on the screen – is the brain of any game. And with each new engine, Carmack’s achieved a higher level of immersion and realism. He’s a hero among coders for particularly elegant programming that pushes the limits of hardware.

In 1991, coding a game called Hovertank, Carmack faced a challenge no programmer had yet tackled: how to get a computer to quickly render a three-dimensional world from a first-person perspective. Previous games wasted processor power by having the computer draw all the walls within range of a character, whether they’re in his field of vision or not; Carmack’s breakthrough was to instruct the machine to draw only what the player would see from his point of view. It was the original first-person shooter. Two years later, Doom introduced variable floor and ceiling heights and walls without 90-degree angles. It was another step toward graphical immersion, giving players the feeling they had been dropped into the game.

With Quake, released in 1996, Carmack went to the next level. Another first-person shooter, it pioneered three-dimensional polygonal characters as well as a more fluid 3-D world that let players see in any direction. “I would spend time just looking down at a corner inside the game,” Carmack recalls, “just walking around, feeling the world is solid, it’s really there.” Quake III Arena added further refinements, including curved surfaces and colored lighting.

For years, games have been racing to catch up to the visual standards of animated films. Before long, Carmack says, game graphics will rival Monsters, Inc. in their detail. When that happens, technical advances in games will proceed at Hollywood’s more measured pace – incrementally instead of in great, creative leaps. Innovators will focus on optimizing existing code, and major revisions will happen less frequently. In effect, Carmack will be obsolete. “There’s a real chance that the next-generation rendering engine will be a stable, mature technology that lasts in more or less its basic form for a long time,” he says. “Programmers will move from being engine coders to being technical directors in the Pixar style.”

Eventually, Carmack says, real-time rendering will be so dynamic that animators will be able to produce films using game engines. Motivated modmakers will have the tools – for free, if Carmack has his way – to bring to life a vision as compelling as the new film Finding Nemo (see Swimming With Sharks). In his book Pattern Recognition, William Gibson writes about a “Garage Kubrick.” Carmack foresees a Basement Disney.

The ideal Carmack has always had in mind is the Holodeck, the immersive simulation device on Star Trek: The Next Generation. It’s science fiction, of course, but a major influence on his thinking all the same. “When I create a game, I’m not telling a story,” he says. “I’m creating an environment in which interesting things will happen.” If, as Carmack believes, graphics engines are reaching a plateau, “now is the time to do a generalized environment” that would be the ultimate mod – a programmable virtual reality like the Metaverse described by Neal Stephenson in Snow Crash.

Carmack presented the idea to id in early 2000, not long after he finished Quake III Arena, but all he got was blank stares. His colleagues wanted to make Doom III, a sure-fire hit, not some futuristic environment the market might not embrace. He admits to some disappointment, but he’s not done with the idea. “It’s a moral imperative that we create this,” he says.

In the meantime, there’s business to attend to. The graphics advances Carmack is pioneering with Doom III are already attracting attention from the software developers who license id’s code. Carmack’s engines are the gold standard of the industry; the engine for Quake III fetches a $250,000 flat fee plus a royalty from the licenser. License deals now account for 20 percent of id’s revenue, according to CEO Todd Hollenshead. That means game sales still bring in most of the money – but competition there is growing. In the past three years, shooters like Halo and Unreal Tournament have eaten into Quake III Arena’s market share. But id could pocket some cash from Microsoft. Redmond keeps calling, trying to convince the company to release a version of Doom III for the Xbox: “We’re being offered a pretty significant amount of money to sit on it until an Xbox port is done,” says Carmack. id hasn’t announced a decision yet.

Late one night in September 2000, Carmack saw the light – everywhere. He was months into the development of Doom III, and as he sat in his office, he watched neon splash the Mexican restaurant across the street. Starbursts refracted on the hood of his Ferrari in the parking lot. A beam from a streetlamp shot through the slits in his blinds and onto the lizard statue resting on a shelf of thick technical books.

Carmack had long perceived the world through an Impressionist lens, with light being key to perception. Now he was beginning to see realistic, dynamic lighting as the last remaining roadblock between the current state of game graphics and the Pixar-quality engine he envisions.

Dressed in jeans and a white Advanced Micro Devices T-shirt, Carmack sat at his desk and pecked at his PC. Onscreen, a bloodred cube hovered below a matrix of white lamps. As the cube spun, it cast a variety of shadows on the steel-colored walls surrounding it. The interplay looked compelling, Carmack thought, reaching for his soda, but he still had to figure out how to unify the light-surface interactions. More urgently, he needed another diet Coke.

When he rose from his desk, his photosensitivity lingered like a hangover. He was so immersed in his task that he saw the world around him as an optical display. In the shower the next morning, three perfect bars of light reflected on the tiles. Hey, Carmack thought – that’s a diffuse illumination by a specular reflection.

For centuries, artists have tried to capture the effects of light to present an image just as the eye sees it; Monet meticulously painted every spot of sunlight on the London’s Parliament building at different times of the day. Today’s game developers struggle with the same thing. “It’s about creating a suspension of disbelief, and the thing that lets you do that is lighting,” says Andy Thompson, director of advanced technology marketing for ATI Technologies, the Ontario-based company that manufactures Radeon graphics cards. “Lighting is critical to making people think a game is real.”

Real enough to be frightening, anyway. Ever since the Nazis in Wolfenstein 3-D screamed at the player, id has aimed to horrify. Doom III is the creepiest yet. “We want to scare people,” lead designer Willits says. “You do that by surprising people. And you do that by grossing them out.”

To induce that level of fear, Carmack knew he had to eliminate what he refers to as “the Hanna-Barbera effect.” In Road Runner cartoons, he says, you can always tell which boulder is going to fall, because it’s a slightly different hue than the static background. The light doesn’t look right. Until now, lighting effects in games were dictated by graphics cards in a limiting way. Games couldn’t render general-purpose dynamic shadows, Carmack says, so they used light maps, static dark patches essentially painted on a surface.

Inside id’s art room, a dark chamber littered with H. R. Giger alien statues and books about Japanese art, animator Fred Nilsson boots up a rudimentary level of Doom III called Delta Map 2. The goal here, Willits explains, is to find an imprisoned scientist who can help point the way to the gates of hell. In classic id style, the halls leading through the base are haunted-mansion spooky: narrow, winding, and flickering with light. More harrowing, players can inadvertently shoot out the lights, which lures nocturnal mutants onto their path. It’s certainly scary stuff. As we creep past a room where a zombie is munching on the brains of a dead scientist, an Imp flies up through the grates below, shooting fireballs.

When he makes the engine for a new game, Carmack has to predict the capabilities of PC hardware two years down the road. “There’s no real hard and fast rule,” he says, “but roughly we buy the very best stuff we can get at the start of development and target that as what we’re shooting at.”

For Doom III, that meant predicting in 2001 what hardware would be available today. Given the then new generation of programmable graphics cards represented by Nvidia’s GeForce, Carmack figured he could leave static light maps behind. These new cards were powerful enough that coders could script their own real-time lighting algorithms to render, for instance, a moving ceiling fan casting an intermittent shadow on a demon’s hide, or a fluorescent light reflecting off the teeth of a chain saw.

By betting on a certain kind of hardware, Carmack creates industry standards. At stake are millions of dollars in upgrades, not only in graphics cards but in the surround-sound systems gamers will need to hear an Imp creeping up around the corner. According to postings on the Web, ATI was so eager to capitalize on Carmack’s latest engine that the company leaked a Doom III demo on the Net last November. Gamers hypothesized that ATI did it to showcase its high-end Radeon 9700 Pro series before Doom III hits the shelves. ATI’s Thompson denies knowledge of a leak. But he admits the game is already reshaping the $1.6 billion graphics card business. “Doom III shows off the performance of the hardware,” Thompson says, “and that justifies us showing off more new chips next year.”

Even with all that processing power at his disposal, Carmack then squeezes out every ounce of performance he can. For the game’s monsters, he created an algorithm to reduce a 500,000-polygon character to a mere 2,000 when viewed at a distance – just the amount to ensure the game’s speed. Carmack also fixed the tendency for shadows to invert when the viewer’s eye was inside the darkness by calculating the shadows from above instead of from the player’s point of view. The technique inspired several dissertations and a name: Carmack’s reverse.

“It was one of those really elegant solutions,” says Carmack, that could come only after grueling hours of work. “There’s this cultural stereotype of a person staring off into space until a light bulb turns on, but that’s just intellectual laziness. You have to get inside a problem and work it.”

“Next victim!” shouts Todd Hollenshead.

Just outside the Mesquite Rodeo, pasty guys in black T-shirts line up next to a small stage, waiting to go under the knife. Wielding clippers that move in flashes of silver, Hollenshead and Marty Stratton, id’s other “biz guy,” cut Quake’s clawlike logo into fanboy skulls.

The occasion is the seventh annual QuakeCon, a gathering where Carmack will unveil a more advanced demo of Doom III for the public. More than 3,000 armchair space marines have made the trip, schlepping along their elaborately configured PCs to compete in a marathon 72-hour deathmatch. The entire convention hall is snarled with cables networking the gamers, who chug free samples of high-caffeine Bawls soda as they play.

Carmack peels into the parking lot in his cherry-red Ferrari and beelines for the rear door. “I never wanted to be a celebrity,” he says backstage. But he braves the crowd to deliver “one of my normal long technical rambles.” What the crowd really wants to know is what Carmack will do next. Will he ever make another engine now that he’s spending so much time over at Armadillo Aerospace? “That’s always the way it is with John,” according to id co-owner Adrian Carmack (no relation). “At any point he might leave.”

For now, Carmack plans to remain close to earth – and id. Doom III will spawn spinoffs: a mission pack with more single-player features, a multiplayer expansion add-on, possibly a program to dynamically render unique new levels on the fly. Eventually, he’ll begin work on his next graphics engine. Doom III won’t be his last, he says, but the days of hammering out a new engine every few years are coming to an end. Soon, he says, “hardcore programmers won’t have a good reason to write engines all the time. They’ll do it because it’s kind of fun.”