Pentium II at 600Mhz alongside Voodoo 3 Emulated on 86Box alongside M6 Mac Mini

Why 86Box Cares About One Core

86Box emulates an old PC at the hardware level. CPU timing, chipset behaviour, ISA and PCI buses, graphics chipsets, audio devices, and disk controllers all matter to getting duration application to behave properly. That does not justify identical achievement to the first hardware, as the Cinebench results below illustrate. Frankly the project is fascinating and the attempt has me in awe, but that accuracy is expensive, and nearly all of the disbursal lands on a sole presenter thread.

The applicable consequence is fairly straightforward in that center figure barely matters here. What really matters is how accelerated one center can run, and how lengthy it can clasp that speed without throttling. Luckily for the Mac Mini current that's exactly anywhere Apple Silicon has been strongest, and anywhere the M6 shines.

The Machines

Every 86Box test uses the identical device configurations, the identical disk images and emulated hardware. The difference set is:

• Mac Mini M6 (12-core CPU, 24GB, this assessment unit)
• Mac Mini M4 (base 10-core CPU, 16GB)

Testing was using a slightly modified build of 86Box 6.0, released on May 31, 2026. The squad improved CPU emulation achievement on ARM hosts and added an ARM64 just-in-time recompiler for Voodoo graphics. That second alter is particularly applicable here, since the Windows 98 device is operating an emulated Voodoo 3. Some credit certainly belongs to the software, Apple Silicon has accelerated cores, but 86Box is additionally getting improved at using them.

I have not measured the uplift from an older 86Box version, stated this is a assessment of the M6 and not 86Box.

Why 100% Is the Only Acceptable Number

86Box reports an productive emulation speed as a percent of its mark speed. 100% method the presenter is keeping gait alongside the emulator's timing model, so item small is a genuine problem. It does not justify that a benchmark volition mark exactly as it would on a bodily CPU at the identical clock.

Consistency additionally matters additional than the average here. Even concise dips create audible artefacts since audio hardware is fed in genuine period and starved buffers are heard as concise dropouts which are irritating. Basically any 86box setup that regularly dips below 100% volition not awareness right, and having adequate headroom on the scheme to comfortably emulate the mark speed volition be a much additional stable experience.

The outcome is that turns all presenter device into a ceiling fairly than a score, and for any stated emulated configuration there is a maximum CPU clock the presenter can prolong at a flat 100%. Because nearly all of that activity falls on one thread, single-threaded achievement moves this ceiling substantially, which is the entire logic the M6 is engaging for this.

Test Method

To discover the ceiling, I made a tradition build from the identical commit as the 6.0 publish (build 9001), extending the frequence tables in 50MHz steps up to 800MHz. The Deschutes frequence array patch is accessible if you desire to try it yourself against that tagged release. It adds 500–800MHz entries alongside recollection and cache timings scaled to keep about the identical admission latencies, and keeps the AT bus at 8.33MHz. The emulation code is alternatively unchanged. Both Macs used this build for the extended tests.

Some power contend a Pentium II at these speeds is not era appropriate. I contend it is fair a small overclock!

The emulated device is alternatively fixed throughout all run:

• Slot 1 motherboard alongside Pentium II (Deschutes), clock varied per run
• 256MB of memory
• Voodoo 3 emulated VGA alongside 16MB of video recollection (2 threads)
• Windows 98 SE

The burden is deliberately a small awkward, Cinebench 2000 operating its CPU test during Winamp 2.76 plays a 16-bit 44,100Hz PCM WAV in the background. The audio is a bit of an achor as it's a real-time person of the emulated hardware, so any instant it falls rearward is immediately audible.

The continue circumstance is ultimately subjective but strict. If I comprehend a dropout, or the reported emulation speed falls below 100% at any point, the run fails.

Results

The basis M4 Mac Mini holds 500MHz, alongside the two Macs extremely stable throughout the complete Cinebench 2000 and 3DMark 2000 SE runs at that speed. At 550MHz, the M4 starts to hitch. They are minor interruptions in Cinebench, but additional noticeable during the 3DMark demo, and adequate to neglect the run.

The M6 passes 550MHz and 600MHz, which is incredible. At 600MHz it held a flat 100% through the two Cinebench 2000 runs and the 3DMark 2000 SE demo, the second giving it 7–8 minutes of uninterrupted testing. That makes the highest passing clock 20% higher than the M4's in this setup.

The CB2000 scores, for anyone who cares:

Those scores depict the emulated CPU at all clock. A completed render is not adequate to pass, the 10.08 CB outcome at 650MHz came alongside one or two audible underruns. In reality, I think I may be being too resolute on that run, and backdrop action could have caused the hitches. But staying true to the methodology, 600MHz is the result and leaves a defeat of headroom. The 800MHz choice is there to broaden the test range but plainly not a speed either Mac can sustain.

There is an engaging wrinkle whenever comparing these scores alongside genuine hardware. An Ars Technica forum thread collecting Cinebench 2000 results includes the following user-reported figures:

Our emulated 450MHz Pentium II scores 7.02 CB, concerning 61% higher than that bodily PII 450MHz result. At 300MHz the gap is larger still, alongside 4.62 CB against 2.38 CB. Bizarrely, our 600MHz outcome of 9.28 CB lands nearly exactly alongside that 800MHz Coppermine. These are idiosyncratic forum submissions from distinct systems, so they provision duration environment fairly than a controlled comparison, but the discrepancy is substantial. It's additionally value a citation this benchmark type is extremely old and lengthy before it became the famous benchmark it is today.

I do not yet cognize why, it's imaginable recollection bandwidth and cache timing inside the emulated device power have item to do alongside it. There is several applicable former in 86Box's v3.0 publish notes, which explain that P6 emulation was not completely exact since of the complexity of out-of-order implementation and L2 cache behaviour. Deschutes timings were tuned to get fairly near to genuine hardware. But yeah, not sure.

The discrepancy is already current at 300MHz and 450MHz, below the entries added by my patch. For this review, 600MHz remains the highest passing environment in this 86Box configuration, alongside a 20% higher stable clock than the M4. The CB2000 scores are helpful for showing how that configuration scales, but I would not use them to assertion equal achievement throughout application on a genuine Pentium II or Pentium III.

Even the M4's 500MHz is notable alongside the overclocked 9950X3D demonstration I was using for context. I do not have a matched x86 run on this tradition build, but the basis M4 was already a much additional capable retro device than the first 450MHz bounds let me establish.

The video below is from the before 450MHz run on the M6, anywhere OBS was additionally compositing and encoding 720p30 H.264. It documents that run, fairly than the new 600MHz result.

What the Host Is Actually Doing

86Box parks the activity on two 'super' cores, apparent as cpu6 and cpu7, alongside the remainder of the bundle mostly idle. On the M6 those two cores average concerning 4,483MHz during the 450MHz Cinebench 2000 run and concerning 4,710MHz during the 600MHz run, peaking at 4,788MHz. Core action climbs alongside the emulated clock too, from approximately 41% on all occupied center at 450MHz to 45 to 49% at 600MHz, so there is motionless genuine headroom remaining at the highest passing clock. Whole bundle use never passes 26%.

The M4 operating the identical configuration keeps its occupied cores nearer to 3.7GHz, and that gap is apt most of the logic it stops at 500MHz.

Where This Leaves the Mini

For this particular use, the cheapest new Mac desktop is an unusually fine machine. The basis M4 already gives me a stable 500MHz Pentium II, the M6 takes that to 600MHz, alongside intact audio and a complete 3DMark demo at 100%. That is an outstanding outcome from a small, silent box under €1,600. It is the clearest case I have established so far anywhere the M6's single-core guide translates into item you can really awareness fairly than item you peruse off a chart.

Looking at the benchmark numbers, and anecdotal evidence accessible online from era appropraite benchmarks, this places my virtual scheme in Pentium III territory. One day I'll appearance to try XP emulation in 86box.

Power and ThermalsOn clasp until powermetrics reports M6 CPU bundle power. Check back.

Gaming BenchmarksCyberpunk 2077 at native 1080p, anywhere the 12-core GPU takes over.