Game Cartridges

A cartridge holds ROM chips wired directly to the console’s address and data bus. The CPU fetches instructions from them exactly as it would from internal ROM, so there is no load: switch on, and the first instruction is already there.

That is the obvious advantage, and it is not the interesting one. The interesting one is that the cartridge is on the bus, so it can be more than memory. Anything the console’s designers left out can be added later by putting it in the slot — extra RAM, a battery, bank-switching logic, a second processor, another sound chip. A cartridge console is a machine whose specification is not finished at launch.

What a cartridge contains

• Program ROM, mapped into the CPU’s address space — on the NES, $8000–$FFFF.
• Character ROM or RAM on machines whose graphics hardware fetches tiles itself. The NES’s PPU reads its pattern tables straight off the cartridge, so the graphics data never passes through the CPU at all.
• A mapper: logic that swaps banks of ROM in and out to beat the address space. The NES’s 6502 can address 64 KB total; Kirby’s Adventure is 768 KB.
• Save hardware — static RAM with a lithium cell. Nintendo’s specifications name the CR2032, still the cell in a bedside clock today, and the reason a 1987 cartridge can hold a save for decades and then lose it in a week.
• Enhancement silicon — the Super FX in Star Fox, the SA-1, the DSP series; on the NES, mappers such as MMC5, Konami’s VRC6 and Sunsoft’s 5B added sound channels the console itself did not have.

The constraints that shaped games

• Cost. Mask ROM is expensive and must be ordered months ahead, in quantity, with no way to correct a mistake. Publishers guessed print runs and ate the difference. This is why Japan got the Famicom Disk System, and why cartridge-era games were finished before they shipped rather than patched afterwards.
• Capacity is bought, not given. Every extra kilobyte costs money on every unit sold, so compression, tile reuse and shared assets were commercial decisions as much as technical ones.
• Nothing can be fixed. A bug in a cartridge is permanent. It is a different discipline from disk or download, and it shows in how conservatively the best cartridge games are engineered.

Cartridges as control

The slot was also a business instrument. Nintendo’s lockout chip meant a cartridge had to answer a challenge from the console before it would run, which made unlicensed publishing a legal and technical problem, not just a commercial one. Combined with Nintendo manufacturing every cartridge itself, it gave the platform holder control over what reached shelves — the mechanism that ended the shovelware glut after the 1983 crash, and the model every console since has followed in some form.

📘 Nintendo went further than the hardware. Its Game Cassette Registration Specification, dated 18 February 1987, set aside a fixed area near the top of the address space where every cartridge recorded a maker code, a registration number, its cassette type and checksums over the program and character data — a manufacturer’s registry, embedded in the ROM of every game. (That document is an internal Nintendo specification, described here and not reproduced.)

Beyond the console

• Commodore 64: the expansion port takes ROM cartridges that autostart by placing a recognised signature at a fixed address — and also freezers, fast-load hardware and RAM expansions. Same principle, applied to a home computer.
• Sega Mega Drive: straightforward ROM at first; later titles added bank-switching hardware to break past the initial ceiling, and the Virtua Racing cartridge carried its own DSP.

See also

• Bank switching — how mappers beat the address-space limit.
• MMC1 and MMC3 — the NES’s most-used mappers.
• NES — the machine that made the cartridge a platform.
• Memory management — the address-space problem in general.