A Mac Mini built around Apple's M6 processor sustained a 600MHz emulated Pentium II configuration in a test of 86Box, demonstrating the headroom available for detailed retro-PC emulation on a modern Arm-based desktop. The result came from a reviewer using a modified build derived from 86Box 6.0 rather than a standard physical Pentium II system.
86Box models a period PC at the hardware level, including processor timing, chipsets, buses, graphics, audio and storage controllers. That approach aims to reproduce the behavior expected by older software, but it also places a heavy computational load on the host. In the review's setup, most of that work depended on one host thread, making sustained single-core performance more consequential than the total number of processor cores.
The tester extended the emulator's Deschutes processor frequency table in 50MHz steps up to 800MHz while otherwise retaining the same emulation code. The Windows 98 configuration included an emulated Voodoo 3 graphics card. The 86Box 6.0 line also brought improvements for Arm hosts and an Arm64 just-in-time recompiler for Voodoo graphics, meaning both hardware and software contributed to the outcome.
The pass condition was deliberately strict. Cinebench 2000 ran while Winamp 2.76 played an uncompressed 16-bit, 44.1kHz audio file. A clock setting failed if reported emulation speed fell below 100 percent or if an audible dropout occurred. Under that method, the M6 machine passed at 550MHz and 600MHz, including two Cinebench runs and a seven-to-eight-minute 3DMark 2000 SE demonstration at 600MHz.
The same test placed the base M4 Mac Mini's stable ceiling at 500MHz. It began to hitch at 550MHz, giving the M6 a 20 percent higher maximum passing clock in this particular configuration. The M6 completed a 650MHz render, but audible underruns caused that run to be rejected. Neither Mac sustained the available 800MHz setting.
The reviewer cautioned against treating the displayed clock or Cinebench score as a direct claim about equivalent physical hardware. Emulation timing, memory behavior and cache modeling can produce results that differ substantially from a real Pentium II or Pentium III. The useful finding is narrower: under one repeatable workload, the M6 host kept a complex emulated PC at full target speed through tests that exposed even brief timing failures.
That distinction matters for retro-computing users. Smooth emulation is not simply about completing a benchmark; real-time audio and graphics must remain synchronized. The experiment therefore measures a practical stability ceiling for this setup, not a universal ranking of old and new processors.
