How the Estimate Works
The test runs in two stages. First, a small file (this site's own favicon) is fetched several times in a row, each with a unique cache-busting query parameter so the browser can't serve a cached copy, and the round-trip time of each request is timed with performance.now(). The first sample is discarded to avoid counting DNS/TCP/TLS connection warm-up, and the median of the remaining samples becomes the ping/latency result.
Second, a few larger static files already published on this site are fetched back-to-back (also cache-busted), and the total bytes transferred divided by the total elapsed time gives an effective throughput, converted into megabits per second (Mbps = bytes × 8 ÷ seconds ÷ 1,000,000). A progress bar tracks each chunk as it completes.
Why Same-Origin Only
Most speed test sites route your test traffic through a purpose-built third-party test server (or a network of them), chosen to be as close to you as possible, specifically to measure your raw internet connection. This tool takes a different, intentionally simpler approach: every single request goes to this site's own server, using files that are already published here for normal site use.
That keeps the test dependency-free (nothing to fail if a third-party test server is down or blocked) and transparent about what's actually being measured, but it does mean the number reflects the path between you and this server specifically, including this server's own response speed - not a dedicated, optimized speed-test endpoint. That's exactly why it's labeled an approximate estimate rather than a definitive result.
Common Use Cases
Reading the Results
| Metric | How It's Measured | What It Tells You |
|---|---|---|
| Ping (ms) | Median round-trip time of several small cache-busted fetches | Lower is better - reflects responsiveness/latency to this server |
| Download (Mbps) | Total bytes of larger files ÷ total transfer time | Higher is better - an approximate throughput estimate to this server |
| Data downloaded | Sum of all bytes fetched during the download stage | How much data the estimate is based on |
Because everything is same-origin, running the test again immediately after can sometimes read faster due to warmed-up TCP connections or intermediate caching layers outside your browser (like a CDN edge cache) - run it a couple of times and consider the results together rather than relying on a single number.
Frequently Asked Questions
The first request typically includes DNS/TCP/TLS connection setup overhead, which would skew a "latency" reading upward. Discarding it and using the remaining samples gives a more representative number for steady-state responsiveness.
A median resists being pulled up by a single slow outlier sample, such as a brief moment of Wi-Fi interference, so it reflects typical latency better than a straight average would if one sample happened to spike.
Mbps (megabits per second) is the conventional unit ISPs advertise; MB/s (megabytes per second) is what file-transfer progress bars usually show. Since a byte is 8 bits, the Mbps figure is roughly 8 times the MB/s figure for the same transfer - this tool shows both so you can compare either way.
Warmed-up TCP connections and caching layers outside your browser, such as a CDN edge cache, can make a same-origin repeat run read faster than the first. That's a same-origin quirk of how the test is built, not necessarily your connection actually improving.
Without a unique query string, the browser or an intermediate cache could serve a previously stored copy of the file instantly instead of performing a fresh network transfer, which would make the timing meaningless. The cache-busting parameter forces a real transfer every time.