NETWORK GUIDES
How to read your internet speed test results
Understand Mbps, download and upload speed, ping, loaded latency and why some speed test results are inconclusive.
What the test measures
A browser speed test transfers data between your device and a measurement server. Download speed describes received data; upload speed describes data acknowledged by the server. Both depend on the whole path: the device, browser, Wi-Fi or Ethernet, router, internet connection and server. One test cannot identify which component caused a slowdown.
Mbps and MB are different
Speed is commonly expressed in megabits per second (Mbps). File sizes often use megabytes (MB). Eight bits make one byte, so 80 Mbps corresponds to 10 MB per second before overhead. Actual downloads also include protocol overhead, congestion and application limits. This tester uses MiB internally for binary byte budgets; 1 MiB is 1,048,576 bytes.
Why a higher download number is not the whole story
Downloads, streaming and large files use downstream capacity. Backups, sending files and video calls also need upstream capacity. Interactive tasks depend on responsiveness, especially while someone else is transferring data. Compare download, upload and idle versus loaded latency rather than treating one number as an overall quality certificate.
A blank result is different from zero speed
Network Speed Tester checks whether enough verified data and sustained observation time were collected. If a data budget ends too early, a connection stalls or samples are inadequate, a metric may remain unavailable. That means insufficient evidence, not a measured speed of zero. A brief raw transfer spike is not presented as sustained capacity.
Get a more useful comparison
Keep the same device, browser, test settings and measurement endpoint. Start with a wired connection where practical, then repeat over Wi-Fi from the location you care about. Pause unrelated transfers for a baseline, and deliberately add them only for a comparison. Repeat several times and retain the exported results. A single result is a snapshot, not an ISP service-level determination.
A real measured example — 15 October 2026
In a Windows Chromium engineering run, the six-second capacity phases reported 752.1 Mbps download and 62.0 Mbps upload. The method was capacity-window-1.2+pressure-staircase-1.0. There were 16 idle probes, 39 download-loaded probes and 28 upload-loaded probes. The transfer allowance was 4,096 MiB, and the run also exercised all eight synthetic activity types. These are one connection’s observations, not expected speeds for visitors. The access-link type was not recorded, so this is not a wired-versus-wireless comparison.
Read the uncertainty alongside the number
That example recorded browser scheduler delay above 50 ms at p95. Both throughput windows passed their sample gates, but local scheduling may have affected RTT. A qualified transfer window is not independent accuracy calibration. The example’s seven assessable activity types showed contention against the combined offered workload; hosting inbound reachability remained unverified.
Limits of this public beta
Measurements are not yet independently calibrated against reference links across the supported browser and hardware matrix. Browser timing and the selected server influence results. Do not treat synthetic workload results as certification of a game, streaming provider or conferencing service.
Download the annotated example’s measured fields (JSON). Network identity, session identifiers and credentials are excluded.