VPS Benchmarks (Sep 2026)
We organize publicly verifiable VPS benchmark data so you can compare providers faster.
Data Notes
Snapshot date: 2026-09-22. The board prioritizes public benchmark results and publicly available plan data, and some rows cite public Yabs.sh output directly. Note: this snapshot is not contemporaneous with the /rankings editorial scores; always verify pricing and availability on provider sites.
How to Read These Results
Back to Rankings
Use rankings when you want a faster shortlist by score and value.
View Rankings →Estimate Resources
Use the calculator when you are still unsure about CPU, RAM, storage, and bandwidth.
Use VPS Calculator →Read Review Articles
Use review articles when you want to validate discounts, experience, and provider fit.
Read Reviews →📐 How to Read the Metrics
🧮 Geekbench 6
Overall CPU performance: single-core scores drive serial tasks like web requests and compilation; multi-core reflects parallel throughput.
💿 FIO 64K Random I/O
Disk performance: 4K/64K random reads and writes are the real bottleneck for databases and hosting — where NVMe shines most.
📡 iPerf3 Throughput
Measured bandwidth: actual transmit/receive rates to the test endpoint. Providers that throttle below advertised bandwidth show up here.
🌏 Datacenter Info
The facility and route each test VM ran in. Performance can vary sharply between locations of the same provider — match against where your users are.
❓ Benchmark FAQs
Where does the benchmark data come from?
A unified test script runs against each provider within the same time window: Geekbench 6 takes the median of 3 runs, FIO and iPerf3 record stable values under default parameters. Entries we cannot test ourselves cite publicly available provider data with the source noted. Every data point carries a snapshot date (current snapshot: 2026-09-22).
How is the overall score calculated?
Overall score = Geekbench 6 multi-core × 45% + single-core × 35% + FIO 64K read × 10% + write × 10%. The weighting favors CPU because most hosting and development workloads hit CPU limits first. For database-heavy workloads, look at the raw FIO values instead of the composite score.
Why does my identical VPS plan score differently?
The same plan can land on different host machines and datacenters. Neighbor load, overselling policy, and test timing commonly cause fluctuations within ±15%. Focus on magnitude differences (30%+) rather than single-digit gaps.
Does a higher score mean it is right for me?
Not necessarily. Benchmarks measure raw performance. Routing, price, datacenter location, and support matter just as much — for visitors from China, an optimized route often beats an extra 20% benchmark score.