Testing Methodology
Every host is measured the same way, on the same plan size, with the same automated benchmark — so the numbers are actually comparable.
How we test
For each provider we buy a real plan, deploy our custom benchmarking plugin, and run a benchmark session made up of five individual runs with pauses in between. We repeat that session every 8 hours across a full 24-hour day to even out time-of-day load, and if a host's results come back inconsistent, we keep testing for another full day to properly capture the swing. We also keep an uptime monitor running for the full month of the subscription.
The goal isn't to stress a server until it falls over — it's to measure how each host performs in realistic, standardized conditions, so a score for one provider means the same thing as a score for another.
Every host is tested on a 4 GB RAM server so the playing field is identical. If a provider only sells plans with more than 4 GB of RAM, we try to limit the server to 4 GB using Java arguments and launch options so it competes on the same footing as everyone else. When that isn't possible, we state it explicitly in the review and its detailed information.
What we measure
Our plugin records 140+ data points per run, grouped into categories that reflect the work a real Minecraft server actually does:
- Chunks — generation and loading throughput
- Entities — entity processing under load
- Redstone — tick-heavy redstone logic
- Block updates — large-scale world edits
- Mathematics & multithreading — raw CPU and parallel performance
- RAM / memory — allocation and garbage-collection behavior
- File I/O — disk read/write speed (SSD vs NVMe matters here)
- Virtual threads — modern concurrency handling
How scores work
Each run produces an Overall, Minecraft, and System score, plus a Hardware score that reflects the raw machine you're actually getting. We aggregate the five runs into a single result and report the variance, so you can see how consistent a host is, not just how fast it is on a good day.
Hosts are grouped into performance tiers based on their hardware score:
- Enterprise — 200+
- High-End — 150–199
- Performance — 125–149
- Standard — 100–124
- Budget — under 100
We also rank by Performance per dollar, because the fastest host isn't always the smartest buy.
Support testing
Benchmarks only tell half the story, so we test the humans too. For every host we open a real support request — “help me get a plugin onto my server” — and later try to cancel the subscription. We record response time, tone, whether they actually solved it, and how painful (or easy) cancelling was.
Our principles
- Real data only. We never fabricate or estimate a metric — if we didn't measure it, it shows as N/A.
- Independent. Rankings are driven by the data, not by who pays us.
- Transparent. Any sponsorship or affiliate relationship is disclosed.