Pentium II at 600Mhz with Voodoo 3 Emulated on 86Box with M6 Mac Mini
Review: Retro PC Emulation on the Mac Mini M6 via 86Box
Date: 2026-09-22
Focus: Achieving a stable 600MHz Pentium II emulation.
It appears that the Apple Mac may currently be the premier platform for those seeking cycle-accurate retro PC emulation. Using a specialized build of 86Box (version 6.0), the Mac Mini M6 successfully maintained a stable clock speed of 600MHz.
The Single-Core Bottleneck
86Box operates by emulating an entire PC at the hardware level. To ensure period-correct software behavior, the emulator must meticulously simulate:
- CPU timings and chipset logic.
- ISA and PCI bus communications.
- Disk controllers and sound hardware.
- Graphics chipsets.
"While this level of accuracy is impressive, it comes at a high computational cost, almost all of which is concentrated on a single host thread."
Consequently, having a high core count is largely irrelevant. The only metrics that matter are:
- Single-core clock speed.
- Thermal sustainability (the ability to maintain that speed without throttling).
This is where the M6 excels, as Apple Silicon continues to lead in single-threaded performance.
![Image: A screenshot showing the M6 running a 650MHz Pentium II with Cinebench 2000 and Winamp open]
Hardware and Software Specifications
The testing utilized a modified version of 86Box 6.0 (released May 31, 2026), which introduced two critical updates:
- Enhanced CPU emulation for ARM-based hosts.
- An
ARM64just-in-time (JIT) recompiler specifically for Voodoo graphics.
Host Comparison
| Feature | Mac Mini M6 (Review Unit) | Mac Mini M4 (Base) |
|---|---|---|
| CPU | 12-Core | 10-Core |
| RAM | 24GB | 16GB |
| OS/Build | Modified 86Box 6.0 | Modified 86Box 6.0 |
Emulated Machine Configuration
- Motherboard: Slot 1
- CPU: Pentium II (Deschutes) — Clock speed varied
- Memory: 256MB
- GPU: Voodoo 3 (16MB VRAM, 2 threads)
- OS: Windows 98 SE
The "100% Speed" Requirement
In 86Box, the "effective emulation speed" is expressed as a percentage.
If the value is , the host cannot keep up with the emulated hardware's clock. While a benchmark might still finish, any dip below 100% causes audible audio artifacts (stutters/dropouts) because sound buffers are processed in real-time.
Therefore, the host machine acts as a . The goal is to find the maximum emulated CPU clock that maintains a flat 100% speed.
Testing Methodology
To determine the ceiling, a custom build (build 9001) was used to extend the frequency tables.
// Conceptual patch for Deschutes frequency table
// Added entries: 500MHz, 550MHz, 600MHz, 650MHz, 700MHz, 750MHz, 800MHz
// Memory/Cache timings scaled to maintain access latency
// AT Bus locked at 8.33MHz
The Stress Test:
To ensure stability, the system ran Cinebench 2000 (CPU test) while Winamp 2.76 played a 16-bit 44.1kHz PCM WAV file. If a single audio dropout occurred or the speed dipped below 100%, the run was marked as a Fail.
Final Results
The base M4 Mac Mini peaked at 500MHz. While it handled Cinebench well, it struggled during the 3DMark 2000 SE demo. The M6, however, held a perfect 100% through both Cinebench and the 7–8 minute 3DMark demo at 600MHz—a 20% increase over the M4.
Cinebench 2000 Performance Table
| Emulated Clock | CB2000 Score | Mac Mini M6 | Mac Mini M4 |
|---|---|---|---|
| 300MHz | 4.62 CB | Pass | Pass |
| 350MHz | 5.34 CB | Pass | Pass |
| 400MHz | 6.16 CB | Pass | Pass |
| 450MHz | 7.02 CB | Pass | Pass |
| 500MHz | 7.73 CB | Pass | Pass |
| 550MHz | 8.54 CB | Pass | |
| 600MHz | 9.28 CB | Pass | |
| 650MHz | 10.08 CB |
Note on 650MHz: The M6 achieved a score of 10.08 CB, but it was disqualified due to one or two audible underruns. While these might have been caused by background OS activity, the strict methodology requires a "Pass" to be flawless.
Conclusion: The M6's ceiling is 600MHz, providing a stable experience with a small amount of headroom. The 800MHz option exists in the software but is currently beyond the reach of either Mac.