Benchmark Methodology (How We Test)

How we measure VPS performance consistent, transparent, and fair. No marketing numbers; just reproducible data you can rely on.

VPS Eligibility

We only test VPS plans that meet the following criteria to ensure results match what you would get if you ordered the same VPS today:

  • Specs: Minimum 1 vCPU, 2 GB RAM, 25 GB storage
  • OS: Clean Ubuntu LTS (18.04, 20.04, or 22.04)
  • Billing: Hourly preferred, monthly acceptable
  • Location focus: Indonesia & Singapore (primary), then US & Europe
  • Price range: Up to $30/month
  • Access: Full root access, no provider-side throttling
  • Plan type: Publicly available, not private/beta hardware

Benchmark Stack

We run a mix of synthetic and real‑world benchmarks to cover CPU, memory, disk, and network layers.

CPU & Memory

  • Geekbench 5 & 6 single‑core and multi‑core scores
  • UnixBench overall system performance
  • PassMark CPU Mark & Memory Mark

Disk I/O

  • YABS (fio) IOPS and throughput across block sizes
  • bench.sh quick disk I/O and network snapshot

Network

  • iperf3 throughput to multiple regions
  • Speedtest CLI (Ookla) download/upload speeds and latency
  • nws.sh wide‑region network speed test
  • Ping checks RTT snapshots (e.g., Check‑Host, Ping.sx, Bunny.net)

Real‑World

  • WP Hosting Benchmark Tool simulates typical WordPress workloads
  • OTT Restriction Check verifies geo‑based access

Execution Rules

  • Multiple sessions: morning, noon, and night to capture noisy‑neighbor effects
  • CPU (Geekbench/UnixBench): 2–3 sessions × 3 runs each (averaged)
  • Disk & Network (fio, Speedtest, YABS): 3 runs best consistent result published
  • No tuning: out‑of‑the‑box configuration (no sysctl or kernel tweaks)
  • Consistency checks: cross‑compare fio vs dd vs YABS, iperf vs Speedtest, etc.

Reporting & Transparency

  • Published results include provider, plan, price, region, OS, and date/time
  • Raw outputs are stored (e.g., fio/iperf logs, Geekbench JSON)
  • Anomalies (e.g., throttling or burst credits) are highlighted
  • Methodology updates (e.g., GB5 → GB6) are tagged for fair comparisons

Limitations

Benchmarks are informative snapshots, not absolute guarantees:

  • Performance varies over time with resource contention
  • Some providers enforce CPU or disk throttling
  • Network performance changes with peering and congestion

As VPS users, we need reliable numbers to plan workloads. VPSMetrics provides consistent, real-world-tested benchmarks to help you make smarter decisions