VPS providers love to throw around numbers – "fastest CPU," "blazing NVMe," "unlimited bandwidth." But what do these claims actually mean? And how do you separate real performance from marketing fluff?
This guide explains the key benchmark metrics and how to interpret them.
Why Benchmarks Matter
Benchmarks give you objective data about your VPS's performance. They help you compare providers, detect degradation over time, and make informed decisions about upgrades.
According to recent performance tests, VPS plans are now judged on speed, stability, and affordability combined[reference:0]. Providers that oversell their hardware show clear signs in benchmark results – CPU steal time, inconsistent I/O, and performance drops during peak hours.
Key Benchmark Metrics
1. CPU Performance
UnixBench is the most widely used CPU benchmark for VPS. It runs a suite of tests and produces a single index score. A score of 1000 is baseline from a 1990s SPARCstation – modern VPS typically scores 500-2000+.
Recent testing shows Vultr High Frequency leading in single-core performance with scores exceeding 1500, while Hostease cn2 GIA offers good dual-core performance with double the memory[reference:1]. For WordPress sites, single-core performance matters most. For container clusters or Java applications, memory size is often more important[reference:2].
2. Disk I/O Performance
Disk I/O is tested using fio, measuring sequential and random read/write speeds. 4K random read/write is the most realistic test for web server workloads.
NVMe drives typically achieve 400-500 MB/s random writes, while standard SSDs deliver around 200 MB/s[reference:3]. Many VPS slowdowns are caused by disk I/O bottlenecks, not CPU limitations.
3. Network Performance
Network quality is measured through latency (ping), download/upload speeds, and route stability. Speedtest CLI and ping tests are commonly used[reference:4].
Typical latency from US West Coast to China is 150-200ms, Hong Kong is 30-80ms, and Japan is 50-120ms[reference:5]. If your audience is in China, Hong Kong or Japan nodes are preferable.
How to Run Your Own Benchmarks
The YABS (Yet Another Benchmark Script) is the most popular all-in-one benchmark tool:
curl -sL yabs.sh | bash
For disk testing specifically:
fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --bs=4k --iodepth=64 --size=1G --readwrite=randrw --rwmixread=75
What to Look For
When comparing benchmark results, consider these factors:
- UnixBench scores – Higher is better, but single-core vs multi-core matters depending on your workload
- Disk I/O consistency – Watch for large fluctuations, which indicate overselling
- Network latency – Choose data centers closest to your audience
- CPU steal time – Monitor for values above 5-10%, which indicate resource contention[reference:6]
Need help choosing a VPS? Check our recommended VPS providers.