NETWORK GUIDES
Ping, loaded latency and bufferbloat explained
Learn why a fast connection can still lag and how to compare idle ping with latency while uploading or downloading.
Responsiveness versus capacity
Throughput tells you how much verified data moved in an observation window. Round-trip time tells you how long a request and response took. A high-capacity connection can still feel slow if interactive traffic waits behind large transfers.
Compare idle and loaded RTT
Idle measurements provide a baseline before the main transfer. Loaded measurements sample round trips while download or upload traffic is active. A substantial increase is consistent with queueing or contention somewhere along the measured path, but does not identify the router or ISP as the cause. Browser scheduling can also delay the observed response.
What bufferbloat means
Bufferbloat describes excessive latency when queues hold too much traffic. It often becomes noticeable during uploads, backups or competing household activity. This site reports directional loaded RTT and an excess over the idle baseline; it does not inspect queue depth directly or diagnose a particular device from a single browser measurement.
RTT variation is not direct packet loss
HTTP probe failures, timeouts and endpoint quota rejections are different from measured packet loss. WebSocket traffic uses TCP and cannot be labelled as UDP gameplay. Changes in round-trip time can be useful symptoms, but must be interpreted with the transport, sample count and errors shown in the result.
Run a controlled check
First repeat an unloaded baseline with other transfers paused. Then run the same test while a known upload or download is active. Keep the endpoint, device and browser constant. If possible compare Ethernet and Wi-Fi separately. Record whether the tab was backgrounded or the device was busy, because local scheduling can bias timing.
Measured directional latency example
The 15 October 2026 engineering run had idle HTTP RTT p95 of 37.4 ms, download-loaded p95 of 62.9 ms and upload-loaded p95 of 176.8 ms. Relative to the same idle p95 statistic, the increases were 25.5 ms and 139.4 ms. This illustrates why upload responsiveness can matter even when download throughput is high. The run also flagged browser scheduling delay, so these observations cannot identify a router fault or establish bufferbloat on their own. This was an idle-versus-loaded observation in one run, not a before-and-after router intervention.
Choose the next test from the evidence
If latency rises during uploads, compare the same workload at a lower offered rate before changing settings. If only Wi-Fi behaves poorly, compare locations and a wired baseline. Router queue-management changes depend on the hardware and configuration; consult its documentation and keep a reversible record. The browser cannot independently prove the fault lies in a specific network layer.
Download the annotated example’s measured fields (JSON). Network identity, session identifiers and credentials are excluded.