How to Test Internet Speed Accurately (Wi‑Fi and Ethernet)

Testing your internet speed sounds simple, but small mistakes easily skew the results. This guide shows how to get accurate numbers—on both Wi‑Fi and Ethernet—then explains what they mean.

TL;DR checklist

• Prefer Ethernet when possible; otherwise test near the router on 5 GHz/6 GHz Wi‑Fi.
• Stop downloads/streams, close heavy apps, and pause cloud backup.
• Disable VPN/proxy, ad‑blocking tunnels, and bandwidth‑shaping tools.
• Test multiple times at different hours; use the median (not the single best run).
• Compare results on phone vs laptop to isolate device vs network issues.
• Check ping and jitter first for real‑time apps; then download/upload for throughput.
• Retest after each tweak (channel change, router move, QoS enabled) to validate improvement.

Step‑by‑step for accurate results

1. Connect via Ethernet (Cat 5e/6+) for a baseline. If not possible, use strong 5 GHz or 6 GHz Wi‑Fi in line‑of‑sight.
2. Reboot the router/modem if uptime > 30 days (memory leaks & buffer saturation can distort results).
3. Disable VPN / Proxy / “Gaming accelerators” — they alter routing and add latency.
4. Pause bandwidth consumers: cloud sync (Drive/OneDrive/iCloud), game launchers updating, streaming tabs, torrents.
5. Close background CPU hogs (video editing, VM containers) that can starve the test thread.
6. Check device power mode: turn off battery saver / low data mode.
7. Run 3 sequential tests (wait ~30–60s between runs) — record ping, jitter, download, upload.
8. Repeat in peak hours (evening) and off‑peak (early morning). Keep median of each set.
9. Optional: Run a test while you start an upload/download in another tab to observe latency under load (bufferbloat).

Tip: If run #1 is noticeably slower but #2/#3 converge higher, the first may have been influenced by warm‑up caches or transient background traffic.

What the metrics mean

A “fast” download with poor jitter still produces choppy calls. Always interpret results as a set—throughput alone is not user experience.

Wi‑Fi vs Ethernet

Wi‑Fi introduces shared medium, interference, band contention, and rate adaptation. If Wi‑Fi is slow but Ethernet baseline is fine:

• Use 5 GHz or 6 GHz instead of 2.4 GHz.
• Pick a clean channel (1/6/11 on 2.4; non‑overlapping/DFS on 5 GHz).
• Shorten distance; reduce obstructions (concrete, metal, aquariums).
• Turn off legacy “802.11b” compatibility if all devices are modern.
• Separate SSIDs for 2.4 and 5/6 GHz to avoid sticky clients.
• Update router firmware; enable band steering only if reliable.

See also: Improve Wi‑Fi Speed (link when available).

Interpreting “slow” or inconsistent results

Typical speed needs by activity

Common mistakes that distort tests

• Testing via old 2.4 GHz in a congested apartment block.
• Leaving cloud photo sync or Steam / Epic downloads active.
• Using a VPN “accelerator” that adds 40–100 ms latency.
• Relying on a single test run (outlier) to judge performance.
• Ignoring jitter and only quoting peak Mbps.
• Comparing different servers (geographically inconsistent).
• Running on battery saver (CPU throttled, NIC powersave).

Advanced accuracy tips

• Use at least one single‑thread test to detect artificial multi‑thread boosts.
• Watch latency under load (start a big upload and re‑measure ping).
• Capture traceroute for unusual routing; keep a baseline snapshot.
• Log results over a week; median & 95th percentile tell more than best value.
• Hash but record public IP / ASN (for personal diagnostics) if privacy policy allows.

Interpreting peak vs sustained speed

Short bursts can inflate peak; sustained speed stability matters for streaming and cloud sync. A smooth graph with minimal dips beats a sharp spike plus troughs. If your speed graph “saws,” suspect Wi‑Fi retries, interference, or ISP shaping cycles.

Troubleshooting decision tree (simplified)

1. Ethernet test OK?
   • Yes → Wi‑Fi layer issue (optimize placement, channel, band, interference)
   • No → Go to 2
2. Peak vs off‑peak difference huge?
   • Yes → ISP congestion / oversubscription
   • No → Go to 3
3. Latency jumps during upload?
   • Yes → Bufferbloat → Enable SQM/QoS
   • No → Go to 4
4. Only one device slow?
   • Yes → Device NIC/driver/power settings
   • No → Modem / router hardware or ISP provisioning

Next steps

• Wi‑Fi optimization (channel planning & placement) — guide coming soon.
• Bufferbloat reduction with SQM — deep dive coming soon.
• Compare ISPs using consistent multi‑day logs.
• Evaluate upgrade (fiber vs cable) if upstream is the bottleneck.

Related guides

• Wi‑Fi Speed Test Guide
• Ping, Jitter, and Latency Explained
• Best Internet Speeds for Streaming and Gaming