The Atari 400 and Atari 800 went on sale in November 1979, and they were the first home computers to hand the hard graphics work to dedicated coprocessors instead of making the CPU do all of it. Both ran a MOS 6502 at 1.79 MHz. Around it sat three custom chips: ANTIC for the display, CTIA for colour and sprites, POKEY for sound and input. The 400 cost $550, the 800 cost $1,000. Nothing else on an American shop shelf that Christmas could scroll a playfield smoothly, and nothing else came close on sound.
- Released: November 1979. Announced at a New York press conference on 14 December 1978, shown at the Winter CES in Las Vegas, January 1979
- Manufacturer: Atari, Inc., a division of Warner Communications
- CPU: MOS 6502 at 1.79 MHz (NTSC) / 1.77 MHz (PAL)
- RAM: 8KB at launch; later units shipped with 16KB (400) and up to 48KB (800)
- Custom chips: ANTIC (display processor), CTIA (colour and player/missile graphics, replaced by GTIA from October to November 1981), POKEY (four-voice sound, keyboard scan, paddles, serial bus)
- Launch prices: $550 (400), $1,000 (800). UK, early 1981: around £375 and £750
- Total sales: about two million Atari 8-bit computers across the main production run, late 1979 to mid-1985
How a games company ended up building a computer
Sunnyvale, 1977. Atari had a hit console on its hands and no idea what came after it, so Joe Decuir, Jay Miner and Steve Mayer were pointed at the problem and told to solve it twice. One machine would succeed the VCS. One would take on the Apple II. Internally they became Colleen, the serious computer, and Candy, the cheap games box, named after two Atari secretaries. That naming convention tells you a great deal about the company in 1977, and nothing at all about the engineering, which was about to be extraordinary.
The men signing off the budget were changing at the same time. Warner had bought Atari in 1976; Nolan Bushnell was pushed out in November 1978, and Ray Kassar, a textile executive from Burlington Industries, took over as chief executive. Kassar wanted character graphics, peripheral expansion and a BASIC, because that was what Apple had. Decuir has described the brief from the engineering side rather differently: the machine had to support 1978-vintage arcade games as well as home computer character and bitmap graphics. Both things at once. That tension is the whole design.
Colleen and Candy became the 800 and the 400, and the numbers originally meant something. 4KB of RAM in the 400, 8KB in the 800. Memory prices then fell before either machine reached a shop, both shipped with 8KB, and the model numbers stopped meaning anything roughly a year into their existence. Nobody renamed them.
Why Atari put three custom chips in a home computer
Jay Miner was the system architect, and what he built was a small division of labour in silicon. ANTIC, designed in 1977 and 1978 by Joe Decuir, Francois Michel and Steve Smith under Miner, is a display processor with its own instruction set. You do not tell it what to draw pixel by pixel. You write it a display list, a short program in memory describing the screen line by line, and ANTIC fetches its own data by DMA and builds the frame while the 6502 gets on with something else. Fourteen playfield modes, six of them character-based and eight bitmapped.
Two features fall out of that design and they are the reason Atari games look the way they do. Display list interrupts let the CPU break in on a chosen scan line and change colours, character sets or pointers mid-frame, which is how a 1979 machine puts more colours on screen than its mode is supposed to allow. And ANTIC has two hardware scroll registers, HSCROL and VSCROL, that shift the playfield by fractions of a character cell. On a Commodore PET or an Apple II, scrolling meant the 6502 shovelling bytes around memory and a screen that jerked a character at a time. On an Atari, it meant writing one number into one register.
CTIA, designed by George McLeod with technical help from Steve Smith, handled colour and sprites. Four players eight pixels wide, four two-bit missiles, hardware collision detection that registered overlaps without the CPU being asked, and 128 colours from 16 hues at 8 luminance levels. From late 1981 it was replaced in production by GTIA, which extended that to 256 colours and added three alternative colour-interpretation modes.
POKEY was Doug Neubauer’s, and it did an unreasonable amount of work for about 5,000 transistors. Four semi-independent audio channels with 8-bit frequency resolution, configurable as four 8-bit voices, two 16-bit ones, or a mixture. It also scanned the keyboard, read eight paddle potentiometers, generated pseudorandom numbers, and drove the serial bus. Its name is simply POtentiometer and KEYboard, which is the least glamorous origin story of any famous sound chip.
Decuir’s verdict on the architecture, years later, was blunt: ‘the thing that Jay did, just kicked everybody’s butt’. He was right, and it is worth being clear about why. In 1979, on a machine selling for $550, Atari shipped hardware sprites, hardware collision detection, a programmable display processor and four-voice sound. The Commodore 64 would not arrive with a comparable set of ideas for another three years.
Why the Atari 800 has no expansion slots
Because of the FCC, and because Atari had nobody to blame but its own industry. The Commission had spent years fielding complaints about radio interference from cheap games console modulators, and by the time the 400 and 800 went through testing the rules were strict. Atari’s answer was brute force: a cast aluminium Faraday cage inside the case. Pick an 800 up today and you feel it immediately, because the thing weighs like a piece of test equipment rather than a home computer. It also made the machines expensive to build.
A cage with a hole in it is not a cage, and an expansion slot needs a hole. So the 800’s internal card slots stayed internal, under a hinged lid, for RAM boards and the OS ROM, and everything else had to go out through a connector that could be shielded. Joe Decuir designed that connector in 1978 and called it SIO, the Serial Input/Output bus. Thirteen pins, daisy-chained, with every peripheral carrying its own device identifier and an in and an out socket so the next device could hang off the back of it. Officially 19,200 bit/s, though third-party hardware later pushed it far higher. Plug in a disk drive and it announced itself and worked. In 1979.
There is a postscript to this that is almost too neat. Decuir later worked on the USB architecture, and he credits SIO as its basis. When patent claims were later aimed at USB, the peripheral bus that exists only because an American regulator would not let Atari cut a hole in a metal box turned up in the argument as prior art.
What Atari told the press on 14 December 1978
The press release from that New York conference opens with a sentence that reads like a company clearing its throat: ‘Atari Inc., a division of Warner Communications Inc. and the nation’s leading manufacturer of sophisticated computer-controlled video games, is entering the personal-home computer industry.’ The 400 was quoted at $500 with 4KB. The 800 was $1,000 with 4KB, an 8KB operating system ROM, a 410 cassette recorder, joysticks and software. Expected ship date, 1 August 1979.
They showed the machines at the Winter CES in Las Vegas the following month. Creative Computing’s report split them exactly as Atari intended: the 400 with a touch keyboard, a cassette interface and educational and entertainment software; the 800 with a full ASCII keyboard, cassette, floppy and printer interfaces, and business packages. By the Summer CES in Chicago that June the 400 had crept up to $550, alongside seventeen educational cassette programs at $40 each.
The 1 August date slipped, and what happened instead is one of the better small stories in Atari’s history. Hand-built pilot units went to a Sears warehouse on 29 August 1979 to satisfy a catalogue deadline, and were then collected and taken straight back to Atari. Jerry Jessop, who was there, put it plainly: ‘These were hand built pilot run units…placed in the Sears warehouse and then immediately returned to Atari.’ The first real magazine advertising ran in November 1979, in Byte among others, at $549 and $999.
Pre-announcing a computer in December for a January show creates a specific kind of panic, which is that you now need software. Atari pulled four of its most senior VCS programmers, Alan Miller, Larry Kaplan, Bob Whitehead and David Crane, onto the operating system. Miller was still proud of it decades later. The BASIC was contracted out to Shepardson Microsystems, where Paul Laughton and Kathleen O’Brien wrote it from scratch after an earlier effort to adapt Microsoft’s 6502 BASIC into an 8KB cartridge stalled.
Why Star Raiders sold more Atari 800s than any advert did
Doug Neubauer designed POKEY. Doug Neubauer also wrote Star Raiders. The man who built the sound chip wrote the game that sold the machine the sound chip was in, which is the sort of thing that only happens at a company small enough for one engineer to do both.
Star Raiders is an 8KB cartridge from 1980: first-person space combat, a galactic chart, a star field that moves the way a star field should, and the whole apparatus of shields and damage and hyperspace that Elite and Wing Commander would later inherit. Byte called it ‘probably the single greatest contributor to the sales of Atari’s 400 and 800 series computers’. Decuir said it drove marketing mad, which you can believe, because Atari was selling the 800 as an office machine at the time.
It also settled an internal argument. The strong reaction to Star Raiders inside Atari is one of the reasons Candy, the cheap games machine, was given a keyboard at all. What it got was a flat membrane panel, spillproof and aimed squarely at children, and it is fair to say nobody has ever enjoyed typing on it. It worked, though, and it is the reason the 400 sold to families as a computer rather than a console.
The decision that cost Atari the 1980s
Here is where the respect stops. Kassar and Warner decided the Atari 8-bit would be a closed system, and backed that decision by threatening to sue anybody who wrote software for it. Bushnell, who had argued the other way before he left, described the position without much diplomacy: ‘I felt that the computer system should not be a closed system, we needed to have third party software developers…Atari was saying that if you write software for the Atari computers, we will sue you.’
The engineers said the same thing at the time and were overruled. Alan Miller: ‘All of us on the project strongly urged senior management to make the Atari 400/800 an open design…Unfortunately, management decided to make it a closed system.’ Apple, meanwhile, was documenting the Apple II down to the schematics and letting anyone build for it.
The postscript writes itself. Miller, Kaplan, Whitehead and Crane, the four programmers Atari had put on its computer operating system, walked out and founded Activision in 1979, and in doing so created the third-party software industry Atari had just promised to litigate against. A company can survive making the wrong call. Making the wrong call and firing the argument out of the building at the same time is harder to come back from.
The hardware kept going regardless. The 1200XL, announced on 13 December 1982 and shipped in March 1983, was meant to replace the 800 and instead broke compatibility with a great deal of 400 and 800 software at a price nobody wanted; it was discontinued that June. The 600XL and 800XL that followed later in 1983 fixed it, and the 800XL became the best-selling machine of the whole line.
What the Atari 400 and 800 cost in Britain
Too much, and too late. Ingersoll Electronics, already the UK distributor for the Atari VCS, took on the computers, and the Atari Owners Club Official Bulletin it published in 1980 told British readers that ‘the Atari Personal Computers will be on the market early in 1981’, ‘competitively priced – at around £375, and £750, respectively’. Ingersoll called them ‘revolutionary personal computers’ and ‘the first true “home” models since they plug into a standard home colour TV’, which was pushing it, given that a family television was exactly what every British home computer of the period plugged into.
By August 1981, with Ingersoll showing both machines at the Microcomputer Show at Wembley, the 400 was retailing at around £345 and the 800 at around £645. That same year, Sinclair launched the ZX81 at £49.95 as a kit and £69.95 built. In 1982 the Spectrum arrived at £125 for the 16K model.
Britain never bought the Atari 8-bit in numbers, and the reason was not the engineering. It was the price. Atari was selling an American middle-class appliance into a market that was buying computers on catalogue credit for under a hundred pounds, and it lost that argument before it started. Atari International (UK) eventually took the distribution back off Ingersoll in February 1983, by which point the Spectrum had already taken the country.
Why the Atari 8-bit found its real market behind the Iron Curtain
In Poland, in 1985, roughly 5,500 Atari machines were sold through the Pewex hard-currency shops, and on the black market they changed hands for the equivalent of a university professor’s annual salary. By 1987 something like 100,000 Atari 8-bit computers had been shipped into Eastern Europe, and Atari led the home computer market in the GDR and in Czechoslovakia.
This was the 800XL era rather than the 400 and 800, but it is the same architecture, the same three chips, the same ideas from 1979. Where Britain had a domestic industry undercutting Atari on price, the Eastern Bloc had a hard-currency import channel and a scene of programmers who would take whatever they could get and then push it much further than Sunnyvale ever had. Some of the most technically absurd Atari 8-bit demos ever written came out of Poland, on machines bought at a price that does not bear thinking about.
What Atari had in November 1979, and what it did with it
About two million Atari 8-bit computers sold across the main production run, from late 1979 to the middle of 1985. That is a real number and a decent business.
Jay Miner wanted to follow the 800 with a 16-bit machine built around the Motorola 68000. Atari management told him to stay on the 6502 and stop designing new systems. He left in early 1980, spent time at Zymos designing chips for pacemakers, and in 1982 joined a startup called Hi-Toro on two conditions: that it build a computer, and that the computer use a 68000. Hi-Toro became Amiga Corporation. Agnus, Denise and Paula are ANTIC, GTIA and POKEY with five more years of thinking behind them, built by the same architect who was told not to bother.
Look at what the 400 and 800 actually contained: coprocessors doing graphics work in parallel with the CPU, a display processor running its own program list, hardware sprites with collision detection, and a self-identifying, daisy-chained, hot-pluggable serial bus for peripherals. Every one of those ideas is in the machine you are reading this on. Atari had the whole set working in November 1979, and then spent three years telling programmers not to write for it.
