The core proposition here is simple and well-proven: 86Box does cycle-level PC emulation, almost all of it on a single thread, and Apple Silicon's per-core performance has been the best in consumer hardware for years. The M6 extends that lead. What makes this review worth reading is not the conclusion — faster core, faster emulation, obviously — but the rigor of the methodology and the interesting anomalies it surfaces. The author built a custom 86Box 6.0 binary extending Deschutes frequency tables in 50MHz steps up to 800MHz, then defined a strict pass/fail criterion: 100% emulation speed with zero audible audio dropouts while running Cinebench 2000 and Winamp simultaneously. This is not a synthetic benchmark chase. It's a practical ceiling test — what's the fastest Pentium II clock this host can sustain without the experience falling apart? The M4 hits 500MHz. The M6 hits 600MHz. At 650MHz one or two audio underruns creep in, and the author calls it a fail. The Cinebench 2000 discrepancy is the most technically interesting finding. The emulated 450MHz Pentium II scores 7.02 CB against a user-reported 4.35 CB from period hardware — 61% higher. At 600MHz the emulated score of 9.28 CB lands alongside an 800MHz Coppermine. The author is honest about not knowing why, pointing to 86Box's v3.0 release notes acknowledging that P6 out-of-order execution and L2 cache behavior were approximated rather than fully modeled. This is the kind of caveat that separates a good review from a press release. The host-side telemetry adds useful texture. During the 600MHz run, 86Box parks on two performance cores averaging 4,710MHz, peaking at 4,788MHz, with core utilization at 45-49%. The M4's busy cores hold around 3.7GHz. That frequency gap — roughly 27% — maps cleanly onto the 20% emulated-clock advantage, with the remainder likely eaten by memory subsystem and decode differences. Whole-package utilization never exceeds 26%, meaning this €1,600 box has substantial thermal and compute headroom left. Credit also belongs to 86Box 6.0 itself. The team shipped an ARM64 JIT recompiler for Voodoo graphics emulation, which matters directly here since the test machine runs an emulated Voodoo 3. The author acknowledges this but correctly scopes the review to the M6 rather than trying to isolate software vs. hardware contributions — a separation that would require matched testing across 86Box versions on the same hardware. The review's limitations are clearly stated. Power and thermal data are on hold pending powermetrics support for M6. There's no matched x86 comparison on this exact custom build, though the author references an overclocked 9950X3D demonstration as prior context. The Cinebench 2000 period-hardware comparisons are user-reported forum submissions from varied systems. None of this undermines the central finding, but the author flags all of it, which is refreshing. What lands is the practical verdict: for anyone who wants a silent, small-footprint retro PC emulation box, the Mac Mini M6 is now the best consumer option available, and it's not close. A stable 600MHz Pentium II with Voodoo 3 graphics, intact audio, and full 3DMark demos at 100% — from a box under €1,600 that barely notices the workload — is genuinely impressive. The author's sign-off aspiration to try XP emulation in 86Box feels earned rather than speculative.
