MAME raises system requirements, cuts legacy support to stay maintainable

MAME’s latest reset is bigger than a routine cleanup. The project will now require Windows 10 or later for future Windows builds, move from C++17 to C++20, and skip its April release entirely, with the next version expected near the end of May 2026.

That shift lands hardest on the people who have been stretching MAME across older PCs, older Windows installs, and fringe host setups. MAMEdev is dropping support for the 32-bit x86, or i686, recompiler back end, ending compilation on OpenSolaris and other System V UNIX systems, removing old PowerPC host optimizations, retiring the obsolete macOS aueffectutil tool, and stopping use of prebuilt MSYS2 environments that bundled development tools. The Qt-based debugger is also moving from Qt 5 to Qt 6, which will matter to anyone whose workflow depends on the debugger as much as on the emulator itself.

The maintainer message is blunt about why. MAMEdev says the project’s purpose is to preserve decades of software history and document hardware behavior through source code, and that goal depends on keeping the codebase maintainable enough to keep moving forward. The team says GCC 11 is the oldest supported compiler, and warns that older clang 11 and 12 releases have serious C++20 bugs, with clang 13 able to crash on some constructs. In other words, the modernization is not cosmetic. It is the cost of keeping the preservation engine from calcifying.

The baseline has already been creeping up for a while. MAME 0.287 was the latest official release shown on MAMEdev’s release page, and its Windows x64 builds already required Windows 7 or later plus x86-64-v2 CPU features. Arm64 builds had gone further, requiring Windows 10 or later and Armv8.2-A support. MAMEdev had also raised the Windows x64 baseline to x86-64-v2 starting with MAME 0.274, so the new Windows 10 requirement reads less like a sudden break and more like the next step in a long modernization arc.

For users, the immediate tradeoff is convenience. Old machines and older operating systems will no longer be safe homes for the newest builds. For packagers and front-end maintainers, the change narrows the range of supported build paths and makes legacy workflows less central. For MAME itself, the bet is that fewer aging detours will mean more predictable development, less time spent propping up obsolete platforms, and more room to do the work MAME exists to do: accurately preserve the hardware that made arcade and computer history worth saving in the first place.