What are you planning to use? (e.g., Raspberry Pi, PC, handheld console)

A complete non-merged set (including CHDs) is roughly 30–50 GB . Understanding Set Types

If you are setting up an arcade system today, you will likely choose between the standard MAME 2003 core and the newer core. Both rely heavily on the 0.78 ROMset, but there are differences:

: In the context of MAME, ROMs refer to the data from the original arcade game's circuit boards, essentially the games themselves. These ROMs are extracted from the original arcade hardware and then used by MAME to emulate the games.

Every single game zip file contains all the files needed to run that game, including parent files and BIOS files. This is the largest set in terms of storage, but it is the easiest to use because you can delete games you don’t want without breaking others.

It's this combination of broad game support and lean performance that turned the MAME 0.78 ROM set into a standard for emulation on low-power devices. Furthermore, derivatives like MAME 2003-Plus, which is built on the 0.78 codebase, have become the "golden standard" for arcade emulation in many software packages, directly inheriting this legacy.

For most users, a matched with the corresponding mame2003-plus libretro core is the best approach. How to Properly Use the MAME 0.78 Romset

: The legality of ROMs can be a gray area. While MAME itself is legal, the legality of downloading ROMs for games you don't own can be questionable. Many argue that owning a game or having a ROM for a game you own is legal, but downloading ROMs for games you haven't purchased can infringe on copyrights.

A complete 0.78 romset includes all the games that MAME could emulate (perhaps not always perfectly) in 2003. That doesn't include ALL classic games, but probably most 2d titles you'd want to play from the 80s/90s.

I can provide the exact directory paths and optimal core configurations for your project. Share public link