Xunglee / Speed

HTTP latency test.

Your connection. Measured, not guessed.

Test server Xunglee Server

Download—Mbps

Upload—Mbps

HTTP latency—ms

Jitter—ms

Ready when you are

—Mbps

Results measure the connection between your browser and this test server. They can differ from other tests depending on network conditions and server capacity.

Your connection may charge for this data. Press GO only when you want to start.

Ready when you are

History on this device

Nothing is saved automatically. Save a result to keep up to 20 tests in this browser. No IP address or identifiers are included.

Your browser

IP ↗

Using this tool with confidence

How to use it

This is an HTTP round-trip latency test. It times browser requests and responses to this server; it is not ICMP ping. Run it with the tab visible and other network activity paused for a more comparable baseline.

Worked example

If response times are 20, 22 and 60 milliseconds, their median is 22 ms while their mean is 34 ms. The difference helps distinguish a typical response from an occasional delay. The interface shows statistics from actual HTTP samples.

Interpretation and limits

HTTP timing includes browser scheduling and server processing as well as the network. It cannot identify the exact router causing delay. Low latency here does not guarantee the same latency to a game, meeting service or another country.

Understand your results

Mbps and MB/s

Mbps means millions of bits per second; MB/s means millions of bytes per second. Divide Mbps by 8 for decimal MB/s. Download receives data; upload sends it. These results include HTTP and server overhead.

HTTP latency and jitter

The browser cannot perform ICMP ping here. HTTP latency is the round-trip request time, including server processing. Jitter is the mean absolute difference between consecutive latency samples: sum(|r[i]−r[i−1]|)/(n−1). It is not an ICMP or packet-level jitter measurement.

Why tests differ

Wi-Fi signal, Ethernet, congestion, distance, shared hosting load, PHP capacity, browser CPU and protocol overhead affect the result. Proxies/CDNs, HTTP/2 or HTTP/3, compression, disk spooling and buffering can also change it. This is one server and one route, not a certified line-capacity test.

Download counts received payload bytes using a monotonic clock. Upload counts only bytes acknowledged by the server until its response arrives. Calibration is excluded from speeds but consumes the allowance. Aborted or unacknowledged upload data is not counted in the result.