The internal SSD in the M5 Pro Mac mini is very fast, and it stays that way. It posted about 11.2 GB/s read and 11.7 GB/s write, then held above 11.6 GB/s
both ways after more than 15 minutes of nonstop testing.
First impressions
My unit is the M5 Pro Mac mini with an 18-core CPU, 20-core GPU, 64GB of unified memory and a 2TB SSD. Setup came with one small catch: the machine made me upgrade to macOS Golden Gate before I could get going. It was a bit of extra hassle, not a real problem, and everything worked fine once it finished.
Specs
The M5 Pro model has the fastest memory in the new Mac mini lineup: 307GB/s of unified memory bandwidth, compared with 170GB/s on the M6 models. That is about 80% more bandwidth for the GPU, Neural Engine and big AI models.
Spec | M5 Pro Mac mini (tested unit) | M5 Pro options | M6 Mac mini |
|---|---|---|---|
CPU | 18-core | 15-core (5 super + 10 performance) or 18-core | 12-core (2 super, 4 performance, 6 efficiency) |
GPU | 20-core | 16-core or 20-core | 12-core |
Neural Engine | 16-core | 16-core | Dual 16-core |
Unified memory | 64GB | 24GB, 48GB or 64GB | 16GB, 24GB or 32GB |
Memory bandwidth | 307GB/s | 307GB/s | 170GB/s |
SSD | 2TB | 512GB to 8TB | 256GB to 2TB |
Rear ports | 3x Thunderbolt 5 (120Gb/s), HDMI, Ethernet | Same | 3x Thunderbolt 4, HDMI, Ethernet |
Front ports | 2x USB-C (USB 3), headphone jack | Same | Same |
Networking | Wi-Fi 7, Bluetooth 6, 2.5Gb Ethernet (10Gb option) | Same | Same |
External displays | Up to 3 | Up to 3 | Up to 3 |
Starting price (US) | — | $1,699 | $899 |
Specs from Apple's Mac mini tech specs and Apple Newsroom. The Mac mini announced in late August, 2026 and started shipping on September 22, 2026. Main reason I got M5 Pro, not M6, was because I want to run local AI clusters, and it's important to have Thunderbolt 5 connection. I will write another full review on this in later day.
How I tested
I used AJA System Test Lite, a disk benchmark built for video work, which writes and reads large sequential video frames. I ran it on the internal 2TB SSD twice: a normal pass for peak speed, then a continuous run of more than 15 minutes. The long run matters most, because many SSDs slow down sharply once their fast cache fills up. I was also using few other popular benchmark tools including the "Disk Speed Test" from BlackMagic Design and Atto's Disk Benchmark. Both show similar results with AJA's.
Results
The SSD did not slow down under a long load. After 15+ minutes, write speed was within 1% of the first run, and read speed was actually about 11% higher.
Test (AJA System Test) | Read (MB/s) | Write (MB/s) |
|---|---|---|
Standard run | 11,215 | 11,732 |
Sustained, 15+ minutes | 12,428 | 11,620 |
Change | +10.8% | −1.0% |
For context, 11 to 12 GB/s is close to double what today's fastest Thunderbolt 5 external SSDs reach (roughly 5.4 to 6.2 GB/s, approximate).
Why it holds up
I believe the steady numbers come from Apple's own SSD design. Apple builds the storage controller into the chip itself, so the Mac mini's SSD is raw flash managed by Apple silicon rather than a standard off-the-shelf drive. Apple does not publish its cache details, so this part is my reading of the results rather than a confirmed spec.
Integrated controller. The controller shares the chip with the CPU and unified memory, which keeps latency low and lets Apple tune it closely for macOS.
Large SLC cache. Most SSDs write first to a fast SLC cache, then slow down once it fills. On a 2TB drive with lots of free space, that cache can be very large, which would explain why write speed barely moved over 15 minutes.
Headroom from capacity. Bigger Apple SSDs use more flash chips in parallel, so the 2TB model should be faster than the 512GB base model.
Quick Verdict
The M5 Pro Mac mini's internal SSD is very impressive: about 11 to 12 GB/s both ways, with no slowdown after 15+ minutes of sustained load. For 4K and 8K video editing, big photo libraries or large AI model files, the internal drive will rarely be the bottleneck. Pair it with 307GB/s of memory bandwidth and Thunderbolt 5, and this small desktop feels like a real pro machine. The forced move to macOS Golden Gate was a minor bump on day one, not a reason to hold back.
The 30-minutes Blackmagic Marathon
Blackmagic Design Disk Speed Test is normally a quick check that tells you whether a drive can handle a given video format. I did something it isn't really meant for: I let it loop nonstop for more than 30 minutes.
In a small enclosure like the Mac mini, I expected speeds to start sliding at some point, either from heat buildup or from the SSD's fast cache filling up. That never happened. Read and write speeds stayed flat for the whole run, with no gradual slowdown and no sudden drop. The results at minute 30 looked the same as at minute one, at 12000 MB/s write and 11500 MB/s read.
That says a lot about both the SSD and the Mac mini's cooling. If you're offloading a full day of footage or rendering a long timeline, you care whether the drive holds its speed from start to finish, not how fast it is for the first ten seconds.
Why I Like ATTO Disk Benchmark for This
One of my favorite features in ATTO Disk Benchmark is that you can adjust the read and write I/O weights. Most benchmarks test reads and writes separately, but real work rarely does. With ATTO you can set the exact mix, which let me build tests that look more like how people actually use a Mac mini.
Test 1: Mixed Workload (50% Read / 50% Write)
First I set ATTO to a 50/50 read/write split. This is the kind of load you get when the SSD reads and writes constantly at the same time, such as scrubbing a timeline while render files are being written, compiling a large project, or running apps while the system swaps memory to disk.
Result: about 5,700 MB/s in each direction at the same time.
That's roughly 11.4 GB/s of total throughput with both directions busy. Many SSDs slow down a lot when reads and writes compete, and the M5 Pro Mac mini's internal drive didn't.
Test 2: 100% Read and 100% Write, Sustained for 30 Minutes
Finally, I tested the two extremes, which come up regularly in some workflows:
Videographers often do long write-only jobs (dumping card after card of footage) and long read-only jobs (playing back or exporting high-bitrate media).
Anyone using the Mac mini as a local LLM machine constantly reads large model files. Loading a multi-gigabyte model into unified memory, or switching between several, is almost entirely reading. Faster reads mean less time waiting before you can start prompting.
I ran each test continuously for 30 minutes:
Workload | Sustained speed (after 30 min) |
|---|---|
100% Read | 12,647 MB/s |
100% Write | 10,800 MB/s |
Both numbers held for the full 30 minutes. The 10.8 GB/s sustained write is the more impressive of the two. Writes are where SSDs usually give up once their cache fills, and this one didn't slow down. At 12.6 GB/s reads, a large local model loads from disk in seconds.
Stress Test Verdict
Across every long test (the 30-minute Blackmagic loop, the mixed 50/50 run and the long read-only and write-only runs), the M5 Pro Mac mini's 2TB internal SSD behaved the same way: it hit its speeds and stayed there. This is also an argument for choosing a larger internal SSD at purchase. External drives can't come close to these sustained speeds, and the internal SSD can't be upgraded later. If you're a video creator moving huge files all day or you're running local AI models, the internal storage's consistency is one of this machine's best features.
One More Thing
After 30 minutes of nonstop writes to its internal SSD, the M5 Pro Mac mini's top cover reached 116°F (46.7°C) and its bottom cover stayed at a mild 96.2°F (35.7°C). The one surprising number, 150°F (65.6°C) at the bottom vents, turned out to be the heat of the parts inside, not the vents themselves.
Benchmarks tell you how fast a drive is. They don't tell you what that speed feels like in your hand, on your desk, or tucked under a monitor stand. So after our SSD speed tests, we did one more thing: we pointed an infrared thermometer at the Mac mini to see where the heat goes.
How we tested
We ran a continuous write to the internal SSD for 30 minutes. Sustained writes are the worst case for storage heat: the NAND and the storage controller never get a break, unlike short benchmark bursts.
Right after the run, we took surface readings with a handheld infrared thermometer at three spots:
Top cover: the aluminum lid, center.
Bottom cover: the flat underside, away from the vents.
Bottom vents: the ring of openings on the underside, scanned for the hottest reading.
Results
The top ran about 20°F warmer than the bottom, which means the heat is being pulled up and away from the desk surface.
Measurement point | °F | °C | What it feels like |
|---|---|---|---|
Bottom vents (peak) | 150 | 65.6 | Not a true surface reading (see below) |
Top cover | 116 | 46.7 | Warm to the touch, like a mug of hot coffee |
Bottom cover | 96.2 | 35.7 | Barely warm, close to skin temperature |
A top cover at 116°F is warm but comfortable to rest a hand on briefly. The aluminum lid acts as part of the cooling system, spreading heat across a large surface.
The 150°F vent mystery
The vents read about 54°F hotter than the bottom cover around them. Our conclusion: the thermometer was seeing through the holes to hot components inside the machine.
Three things point that way:
The vents take air in. The Mac mini pulls cool room air through the openings on its underside. A grille with cool air flowing across it should read near room temperature, not 150°F.
IR thermometers measure a spot, not a point. A typical handheld gun has a distance-to-spot ratio of around 12:1. From a few inches away, the measured circle is wider than a single vent hole. So the reading blends the hot parts inside with the cooler grille. The internal parts may actually be hotter than 150°F.
Different materials read differently. IR thermometers assume a surface that radiates heat like a matte object. Circuit boards and chip packages behave that way and read close to their true temperature. Bare metal reads low. What the spot lands on changes the number.
The practical takeaway: the vent reading is a useful hint that something inside is working hard during sustained writes. It is not the temperature of anything you would touch.
What it means for owners
Nothing we measured is a warning sign. Even if the internal parts sit around 65°C (150°F) during long writes, that is within the normal range for flash storage, which typically runs comfortably up to about 70°C.
On Apple silicon, the storage controller lives inside the main chip, not on the storage module. Its heat is handled by the same heatsink and fan as the CPU and GPU. The flash chips themselves do less of the heavy lifting.
For everyday use, keep two things in mind:
Don't block the underside. The bottom openings are the air intake. Set the Mac mini on a hard, flat surface, not a rug, a stack of papers, or a soft mat.
Give the top some room. The lid helps shed heat, so avoid stacking drives or other devices directly on it during long exports, backups or large file copies.
Bottom line
Under 30 minutes of the hardest storage work we could throw at it, the M5 Pro Mac mini stayed warm on top, cool underneath, and never hot anywhere you'd touch. The scary-looking 150°F at the vents is the machine's insides doing their job, seen through the grille. Keep the underside clear, and the chassis temperatures give no reason to worry. Very well done! Apple!
Buy from Amazon
M5 Pro Mac mini
Note: ReviewTonight is an affiliate partner with Amazon. When you click a link and make a purchase, we may receive a small payment, which helps us keep the site running.