The NES Internals Series: Chapter 4 - Game Cartrdiges & Mappers

Introduction

If you haven’t checked out my PPU and CPU writeups, I recommend you go check them out before reading this part, as I’m building on top of things I explained there.

Terminology

  • ROM - a type of memory (which unlike RAM) its data cannot change (this is why its also called ROM - “read only memory”. as you can only read data from it). The data in it is set once by the manufacturer, and cannot be changed again by the CPU.

What is a cartridge?

Modern consoles have full operating systems, and some firmware to boot the operating system (like BIOS or UEFI). The NES doesn’t have any of that - it uses cartridges. These are essentially just memory chips (usually ROM, but sometimes RAM) which contains the data of the game. The cartridge is made up of (in order):

Number 1 - PRG

A ROM chip containing the code of the game

Number 2 - CHR

A ROM chip containing the graphics data (AKA the pattern tables!) of the game

Number 3 - Interrupt Vectors

Pretty self explanatory, go check out the CPU writeup if you don’t know what these are. Reset interrupt vector resides in IRQ interrupt vector resides in NMI interrupt vector resides in

When the CPU boots, it performs a ‘reset’ - meaning it jumps to the address stored at the reset vector (the reset vector is stored at $fffc and $fffd). That data at $fffc and $fffd is the address of the reset procedure written by the game programmers, which starts the game.

Conclusion

NOTE: Some cartridges use CHR RAM, so that the game actually changes the contents of the cartridge during gameplay, and not just the nametables

NOTE: Some cartridges also have a PRG RAM chip with a battery for dumping the CPU memory (address range $0000-$0800 in CPU memory map) when exiting the game, so progress is saved.

So overall, pretty simple - just the raw code & data, and a small section which specifies where are the ISRs located.

What is a mapper?

If we go check the memory map of the CPU, we see that the memory map of the CPU is 16 bits, in the range $0000-$ffff - which gives us ($10000 - $4020) / 1024 = (64-16)KB = 48KB of data. Similarly for the PPU, the range is $0000-$1fff - which gives us $2000 / 1024 = 8KB of data.

This is little amount of data. VERY little amount. So how does the NES display some pretty good looking games with many levels such as Super Mario Bros 3, Kirby, Legend Of Zelda, etc?

The answer is………

Mappers! (I bet that one caught you off guard :)

A Mapper is essentially a small circuit that acts as a “middle man” between the NES bus, and the cartridge data we just discussed (PRG, CHR, interrupt vectors).

It lets the game developers create games that use more than just the limited 48KB and 8KB CPU and PPU memory maps, by exposing some ports (registers) which select a different “section” of the data. These “sections” the cartridge data is divided into are called banks.

And so by selecting different banks, the NES components can access more data.

There are a lot of mappers out there (too many if you ask me). Different mappers supports a different number of banks, expose a different set of registers, and have a different mechanism for changing and setting the values of their bank selection registers.

SOME NOTES:

NOTE: Not every mapper exposes bank selection registers - NROM (iNES mapper number 0) provides no “extra” banks. It supports just the basic NES memory maps, so there is no need for bank selection registers - the same banks are used all the time.

NOTE: Remember I said there are a lot of mappers? Well….. It’s kind of true. There are indeed a lot of mappers, but many of them are based off other mappers with some very minor and small modification. The main mappers used, and the ones which most other mappers are based off are: NROM, UNROM, CNROM, MMC1, and MMC3.

The iNES and NES 2.0 format

So in order for your emulator to run a game, it needs to know some stuff like the mapper the game uses (since your emulator implements the mappers - they aren’t included in the binary file on your computer), what types of data that mapper uses? (RAM, ROM, or both), what are the sizes of the data included? is there a use of battery backed PRG RAM for save data?, etc

A common format for storing this information is the iNES format. It’s structure is (roughly speaking) pretty simple:

------
iNES header
------
PRG data
------
CHR data

Now what is NES 2.0 you ask? Well although the iNES format was more than enough most of the time, it could be better (contain data in a better order, provide some more information iNES was missing)

99% of game ROMs I’ve came across support iNES, so in my opinion supporting NES 2.0 isn’t that big of a deal. Although iNES 2.0 is implemented as an extension, and without many additions to the original iNES - not as a completely new format so adding support for it isn’t that hard.

And as always, for more information check out the NESdev wiki on the iNES and NES 2.0