The reviewer built a custom 86Box 6.0 emulator on a Mac Mini M6 (12‑core CPU, 24 GB RAM) to replicate a Slot 1 motherboard with a Deschutes Pentium II processor, 256 MB of memory, an emulated Voodoo 3 graphics card with 16 MB VRAM, and Windows 98 SE. Stability was tested by running Cinebench 2000’s CPU benchmark while Winamp 2.76 played a 16‑bit 44.1 kHz PCM WAV in the background; any audio dropout or emulation speed falling below 100% caused the run to fail. This real‑time audio load ensures the emulator must keep pace with the hardware timing model continuously. At 600 MHz the M6 maintained a flat 100% emulation speed through both Cinebench 2000 runs and the full 3DMark 2000 SE demo, which lasted 7–8 minutes without interruption.
This shows the specific CPU settings in 86Box used to replicate the Deschutes Pentium II processor with Windows 98 SE as described in the emulator setup
The base Mac Mini M4 (10‑core CPU, 16 GB) could only sustain 500 MHz under the same conditions, beginning to hitch at 550 MHz and failing the test. Telemetry showed that 86Box parks the workload on two performance cores (cpu6 and cpu7); on the M6 these cores averaged about 4,710 MHz during the 600 MHz test, with each core active roughly 45‑49 % of the time and total package usage under 26 %. The M4’s busy cores stayed closer to 3.7 GHz, explaining why its stable clock stopped at 500 MHz. Thus the M6 achieved a stable clock about 20 % higher than the M4 in this specific emulation workload.
Cinebench 2000 scores rose with the emulated clock, reaching 9.28 CB at 600 MHz, though these values exceed period‑hardware results reported on forums—for example, a real Pentium II 450 MHz scores around 4.35 CB, while the emulator gave 7.02 CB at the same clock. The reviewer notes that the score gap reflects differences in memory bandwidth, cache timing, and the emulator’s own simplifications, not a true performance equivalence. Pushing the emulated clock to 650 MHz produced one or two audio underruns, so 600 MHz is deemed the highest stable setting for this configuration. The result highlights the M6’s strong single‑threaded capability for latency‑sensitive, single‑threaded workloads like cycle‑accurate PC emulation, but the reviewer cautions that the Cinebench numbers should not be taken as evidence that the emulated system matches a physical Pentium II or Pentium III in real‑world applications.
This image shows the emulator running at 650MHz where audio underruns occurred, confirming 600MHz as the stable maximum speed for this configuration