Open the leaderboard on the 17NAS homepage, and you will see 138 CPUs ranked from flagship EPYC parts all the way down to entry-level efficiency cores. But plenty of people are not sure how to read it correctly: why is the progress bar not linear? Is it fair to put ARM and x86 side by side? My own CPU ranks 58th — is that good or bad? This article explains the reading method, the data source, and the common questions all at once.
1. Where does this leaderboard's data come from?
The 17NAS leaderboardis not a copy-pasted spec sheet — it is measured on real devices:
- A user runs the CoreMark benchmark on their own NAS or mini PC
- Through thesubmission pagethey upload the result (optionally with screenshots, device and OS details)
- A human reviews outliers before the score is added to the ranking
- Each CPU's samples, devices, and operating systems are fully public on its detail page, for example the Intel N100 benchmark page
The complete data rules are in the CoreMark test method and ranking rules. Simply put: ranking = each reviewed CPU's best CoreMark score, in descending order.
2. Why CoreMark instead of Geekbench?
CoreMark is EEMBC's industry-standard integer performance test, focused squarely on CPU core throughput:
- Runs on every NAS system: Synology, QNAP, fnOS, OMV, Unraid — if it can compile C, it can run CoreMark
- A single-thread score reflects core efficiency: ideal for comparing the low-power CPUs common in NAS builds
- The community can reproduce it: anyone can verify with the same method, which is exactly what makes crowdsourced data trustworthy
It has limits too: CoreMark does not test iGPU transcoding, memory bandwidth, or IO. So when you read the ranking, judge it alongside the spec information on the CPU detail page.
3. Why does the progress bar look so top-heavy?
The leaderboard uses a segmented scale:
- 0 to 50,000 points: takes up 25%
- 50,000 to 1,500,000 points: takes up 75%
Why design it this way? Because the CPUs NAS users actually buy cluster in the 30k to 150k range (N100 around 94k, J4125 around 57k, RK3588 around 82k), while a server flagship (the EPYC 7513 at 1.14 million) would squash a linear axis into something unreadable. The segmented scale keeps the differences at the entry level visible while still conveying where the flagships sit.
4. Where does your CPU rank? Quick tier lookup
Based on the current 138 CPUs collected (rankings update dynamically — refer to the homepage leaderboard):
The golden zone for a home NAS is the "mainstream sweet spot": the best balance of performance, power draw, and price.
5. Is it fair to rank ARM and x86 together?
At the score level, yes — CoreMark measures raw core compute, independent of instruction set. The RK3588 (ARM octa-core) genuinely comes close to x86 quad-cores in multithreaded results. That is a fact.
But when choosing hardware you cannot look at scores alone:
- x86 advantages: full Docker image compatibility, mature virtualization, a strong Intel iGPU transcoding ecosystem
- ARM advantages: often cheaper for the same performance, the RK3588 has a built-in NPU for AI tinkering, and domestic NAS units offer great value
In one line: scores set the tier, ecosystem decides the trade-off. When in doubt, look at a head-to-head page such as RK3588 vs N100.
6. Three correct ways to use the leaderboard
- Position before buying: check whether a candidate CPU's tier matches your scenario (see how to choose a NAS CPU)
- Head-to-head comparison: use neighbouring ranks and comparison pages to see the real gap, for example N100 vs N150 differ by about 15% — a small generational bump
- Verify marketing claims: a vendor says "flagship performance"? Look it up on the leaderboard and see which tier it lands in
FAQ
Q: Why do different people get different scores for the same CPU?
Compiler, OS scheduling, and thermal policy all affect the result; a 5-15% spread is normal. The leaderboard takes the best valid score, while the detail page also shows the average and median — for example the N5105 detail page.
Q: What if my CPU is not on the list?
Run CoreMark once and submit your score. After review, your device will appear on the leaderboard and on that CPU's detail page.
Q: Can a CoreMark score be converted into "how many drives or containers it can handle"?
Not directly, but as a rule of thumb: the entry tier handles 2 drives plus 3 light containers without strain; the sweet spot handles 4-8 drives plus a dozen or so containers; above the performance tier you are basically limited by memory and networking. For container and VM planning you can use the Docker Compose editor and the power calculator to evaluate.
Q: How often is the ranking updated?
Updated in real time as submissions are reviewed — every new sample can shift the ranking. That is exactly why the leaderboard is more useful than a static spec sheet.
