An arcade game was its hardware. A home computer ran whatever program was loaded into it, but for most of the 1970s and 1980s an arcade cabinet held a circuit board made for one game. Capcom’s boards from 1984 to 1988 show how far that went. Fabien Sanglard, who tabulated their chips, found that “even titles produced the same year wouldn’t necessarily feature the same chips”, and summed it up: “Even though games did mostly the same thing (move sprites over backgrounds), the hardware had to be re-invented over and over again.”
The history of arcade hardware after that is a history of standards. First the wiring inside the cabinet was standardised, so any board fitted any cabinet. Then the manufacturers standardised their own boards, so a new game was mostly a new set of ROM chips. The more a board was worth copying, the harder its maker worked to stop it being copied, and Capcom’s answer, a battery that wiped the board’s secrets when it ran flat, still decides whether some of its games work today.
Fast facts
- Earliest games: Computer Space (1971), then Atari’s Pong (1972), built from logic circuits with no processor.
- The 1980s board: usually a Zilog Z80, a Motorola 6809 or, from the mid-1980s, a 68000, often with a second processor for sound.
- The standard cabinet: JAMMA wiring, from 1985-86: one edge connector carrying controls, coins, picture, sound and power.
- Platforms: Capcom’s CP System (1988) and CPS-2 (1993), SNK’s MVS (1990), Sega’s NAOMI (1998), among others.
- Later: boards built from console parts in the 1990s, and PC hardware in the 2000s.
Circuits, not programs
The first arcade video games had no processor and no program. Retro Gamer’s Videogames Hardware Handbook describes the early 1970s as a time “when every arcade company was putting out discrete technology videogames, (games that are built entirely out of electronic logic circuitry)”. Al Alcorn built Pong for Atari that way in 1972, and the Handbook records that Nolan Bushnell’s “stringent requirement” for a tiny budget forced Alcorn to invent his own way of putting the score on screen.
Microprocessors changed that in the mid-1970s. The Handbook tells of Jerry Lawson, a Fairchild engineer, who built a driving game, Demolition Derby, in his garage around the Fairchild F8 processor and had it running by early 1975, and notes that Midway’s Gun Fight is “largely considered the first microprocessor-driven arcade game”. A processor meant that a game could be programmed instead of wired. It did not yet mean that one board could run another company’s game.
One board per game
Through the early 1980s the boards were built round cheap 8-bit processors. The Z80 was everywhere. Capcom’s 1942 (1984) used two of them, one for the game and one for the sound, and Sanglard’s table of Capcom boards from 1984 to 1988 moves between the Z80, the 6809, the 68000 and an Intel 8751 microcontroller, with four different Yamaha and OKI sound chips. Some makers reused a design for several games. Williams built Defender on a 6809 board and kept the same design for Robotron, Joust and Sinistar, adding its own blitter chips to move the graphics.
The results varied enormously within a single year. Sanglard sets two 1987 games side by side. Sega’s X Board, running After Burner, had “dual 12.5 MHz Motorola 68000 CPUs” and a Super Scaler graphics chipset “capable of both scaling and rotating up to 256 sprites, over two background layers, and one ‘road’ layer”, with a Z80 driving a 16-channel PCM chip for sound. Capcom’s board for 1943 had “a 6Mhz z80 CPU with a graphics system capable of animating 32 sprites on top of one text layer and two background layers”, and a second Z80 for FM music. Capcom’s planner Noritaka Funamizu put the gap in his own words: “While they were racing in F1 cars, we were driving Hondas.”
JAMMA: one plug for every cabinet
The standard that changed the business was about wiring, not chips. After the arcade crash of 1983, Edge recounted in 2002, “arcade operators were fed up with having to replace one machine by one manufacturer with a near-identical cabinet from another”, each wired differently. In Japan, the manufacturers’ trade association brought the leading companies together, and in Edge’s telling the heads of those companies “created the JAMMA standard during 1985-6”. Joystick, buttons, sound, monitor, coin mechanisms and power all met at one edge connector on the game board, fed by a wiring loom inside the cabinet.
Sanglard gives the detail: 28 pins on each side of the connector, carrying a four-way joystick and three buttons for each player, two coin inputs, start and service buttons, a mono speaker line, the monitor signals and the power supply. Changing a game became a matter of pulling one board out and pushing another in. The standard also fixed what a game could ask of the cabinet. When Capcom gave Street Fighter six buttons a player, three more than JAMMA allowed, it designed a second set of inputs, wired through an extra loom labelled the “kick harness”.
There had been earlier attempts at an interchangeable system. Data East’s DECO Cassette System of 1980, which loaded games from audio cassettes into a standard cabinet, was in Edge’s account the first time the Japanese industry tried a standardised plug-in board. JAMMA was the one that held.
Capcom’s CP System
Capcom took the next step in 1988: a standard board of its own. The Games Machine reported in June 1989 that the company had spent two and a half years and £5.5 million developing a “super chip” that compressed “the power of ten normal arcade circuit boards into two microchips”, so that an arcade owner could swap games more easily. Sanglard lists the aims behind it: lower costs by building the machine once, with new game boards “containing mostly ROMs” that operators could buy separately; stable tools for the programmers; graphics that could match Sega and Namco; and protection against the bootleggers, who had copied Capcom’s boards built from off-the-shelf chips. In Mexico alone, he reports, an estimated 200,000 bootleg boards were in circulation.
The CP System, later called CPS-1, ran a 68000 at 10 MHz for the game and a Z80 at 3.58 MHz for sound, with two custom graphics chips, CPS-A and CPS-B. Its picture was 384 × 224 pixels, wider than the 320 of the Mega Drive and the Neo Geo or the 256 of the Super NES. Forgotten Worlds and Strider came first, in 1988 and 1989, then Final Fight, and in 1991 Street Fighter II. By Sanglard’s count Capcom released more than thirty games on the board between 1988 and 1995.
The suicide battery
Capcom’s protection lived in its custom chips, and it grew stricter. Up to 1991, Sanglard explains, each game’s CPS-B chip had its behaviour fixed in silicon at the factory. The last revision, CPS-B v21, kept it instead in 18 bytes of RAM inside the chip, which had to be powered at all times, so a battery on the board kept it alive while the cabinet was switched off. When the battery ran flat the chip forgot its configuration and fell back to “default” values that no game used. The board was dead, and the battery earned the nickname that describes it: the suicide battery. Capcom offered a replacement service for a time and then stopped.
The CP System Dash of 1992, sometimes called CPS-1.5, added a second one. Its sound processor was a Z80 variant that decrypted its program as it ran, with the key held in battery-backed RAM. The CPS-2 board of 1993, which made its debut with Super Street Fighter II, encrypted the game program itself. A custom processor, compatible with the 68000, decrypted instructions on the fly, again with the key kept in battery-powered memory. Sanglard reports that no bootleg CPS-2 game is known.
The protection outlived the board. Capcom published forty-two CPS-2 games, the last of them Hyper Street Fighter II in 2003, and the scheme was fully worked out only in 2007, when MAME’s programmers found that the keys were not random. Collectors can now revive a dead board: either by patching the game so that it no longer needs the missing settings, which they call “phoenixing”, or by writing the lost bytes back into the chip.
SNK’s cartridges
SNK approached the same problem from the other side. Its Neo Geo MVS (Multi Video System) of 1990 put each game on a cartridge. SNK’s specification lists motherboards with one, two, four or six cartridge slots, the single-slot board “capable of connecting to a JAMMA unit”, so an operator could run several games in one cabinet and change them without a new board. The same games, on differently wired cartridges, ran on the home version, and a memory card carried a player’s progress between the two.
Polygons and console parts
In the early 1990s the arcade’s lead moved to 3D. Sega’s Model 1 board, which ran Virtua Racing in 1992, took three years to develop, Yu Suzuki told Edge, “whereas new graphics hardware usually takes a period of about 12 months”. Edge described the arms race that followed: Sega’s Model 2 against Namco’s System 22, the board that ran Ridge Racer.
The boards that followed were increasingly built from home-console parts. Sega’s ST-V was a Saturn in an arcade board, and Namco’s System 11, made with Sony, was based on the PlayStation. Sega’s NAOMI of 1998 was a Dreamcast with twice the memory and games on ROM boards. It cost half as much as Sega’s Model 3, it drew in other companies’ games, and Edge noted that it supported the latest version of the JAMMA standard, known as JVS.
PCs in the cabinet
By the mid-2000s the custom board was giving way to the PC. Edge reported at the end of 2004 that Taito’s Type X “is essentially a modified PC running an embedded version of Windows XP”. In 2005 Sega abandoned a third NAOMI, explaining that it “had simply fallen behind advances in PC technology”, and replaced the line with Lindbergh, which Edge called “Sega’s PC-based” board.
Sources
- Fabien Sanglard, The Book of CP-System (self-published, 2026 edition; modern book): Funamizu’s quote and the 1987 comparison of Sega’s X Board and Capcom’s 1943 (pp. 19–21); the 1984–88 chip table, 1942’s two Z80s and “re-invented over and over again” (p. 28); bootleg boards (p. 29); the CP System’s launch games and more than thirty titles (pp. 29–32); the design aims, the 200,000 Mexican bootlegs, JAMMA’s 28 pins a side and the kick harness (pp. 37–39); the 68000 and Z80 clocks (p. 41); the 384 × 224 picture against the Mega Drive, Neo Geo and Super NES (p. 73); CPS-B v21’s configuration RAM, the suicide battery, phoenixing and de-suiciding (pp. 123–125); CPS-1.5 Kabuki, CPS-2, its forty-two games, the absence of bootlegs and the 2007 key discovery (pp. 126–129). https://github.com/fabiensanglard/cpsb
- Retro Gamer: Videogames Hardware Handbook volume 1 (2nd edition, 2016), “Fairchild Channel F”, pp. 18–19: discrete technology games, Pong’s budget and scoring, Jerry Lawson’s Demolition Derby on the F8, and Gun Fight as “largely considered the first microprocessor-driven arcade game”.
- MAME source code,
src/mame/williams/williams.cpp(web, secondary), as cited in the 6809 entry: the Williams 6809 board and the games with and without blitters. https://github.com/mamedev/mame/blob/master/src/mame/williams/williams.cpp - Edge issue 116, November 2002, coin-op history feature: the operators’ complaint after the crash; the JAMMA standard “during 1985-6”, its edge connector and wiring loom; the DECO Cassette System of 1980; Model 1, Model 2 and System 22; the ST-V as “a repackaged Sega Saturn”; NAOMI and JVS; Namco’s System 11.
- The Games Machine issue 19, June 1989: Capcom’s “super chip”, its development time and £5.5 million cost, and Forgotten Worlds as the first game.
- SNK, Neo-Geo Hardware Specification (December 1989, revised 1990 and 1991), Software Development p. 1: MVS boards with one, two, four and six slots, the single-slot board connecting to a JAMMA unit, and the memory card.
- Wireframe issue 5, January 2019, pp. 50–53, as cited in the Data East entry: the DECO Cassette System’s audio cassettes.
- Edge issue 9, June 1994: Yu Suzuki interview (Model 1’s three-year development; Virtua Racing).
- Edge issue 144, Christmas 2004: Taito’s Type X as a modified PC. Edge issue 155, November 2005: Lindbergh as PC-based and the abandoned NAOMI 3.
- Edge issue 65, December 1998, and later issues, as cited in the NAOMI entry: NAOMI’s cost against Model 3 and its Dreamcast basis.
- Edge, November 1994 and January 1995, as cited in the Namco entry: Ridge Racer on System 22; System 11, “jointly created by Sony and Namco”.