LumaProbe measures response time before a transfer, while downloading and while uploading. The comparison can reveal queueing delay that affects calls, games and remote work even when headline speed is high.
Distance to the UK server can raise latency outside nearby regions.Uses about 24 MB per completed test.
Run several tests at different times for a representative result.
These are approximate single-device estimates, not guarantees. Other users, applications, routes and service requirements can change the experience. How the estimates are interpreted →
Compare the increase above idle latency in both directions. A large repeatable increase suggests that traffic is waiting in a busy queue; it does not identify which device or network component owns that queue.
The test first records ten small HTTPS round trips while no deliberate bulk transfer is running. It uses the median successful sample as idle latency. It then downloads three data blocks in parallel while continuing to send small requests, and repeats the same concurrent sampling during a six-mebibyte upload. The two loaded medians are compared separately with idle latency.
LumaProbe assigns an explanatory responsiveness grade from the larger increase: A for no more than 15 milliseconds, B for no more than 30, C for no more than 60, D for no more than 100 and E above 100. The thresholds are published so the grade can be understood; it is not an industry certification or a promise about every application.
Start with avoidable background transfers paused so the first run describes the connection under the test’s controlled load. Label Wi-Fi, Ethernet or mobile correctly and use the same device for comparisons. If possible, add an Ethernet run so a wireless queue can be separated from the broadband or router queue.
Then repeat during the ordinary situation that causes trouble—for example while another household device backs up or downloads a game. Do not deliberately saturate a metered mobile connection without checking the data allowance. A full run transfers roughly twenty-four mebibytes plus small request overhead.
A small increase means the route stayed relatively responsive while the deliberate transfer was active. A large download-loaded increase can affect calls and games when someone receives a large file; a large upload-loaded increase can appear during cloud backups, camera uploads or file sharing. Upload queueing is easily missed by tests that measure only download load.
The result helps identify a condition rather than a component. Queueing can occur in the device, Wi-Fi access point, router, modem, provider access network or another point on the route. A repeatable Ethernet result across several runs is stronger evidence than a single Wi-Fi grade.
This is a short HTTPS browser test against one United Kingdom node. Browser scheduling, server distance, VPNs, security software and a very fast connection can change the sample count and measured rates. It does not inspect router queues, identify a particular queue-management algorithm or emulate every type of real-time traffic.
A good grade does not prove that a game server or video platform is healthy. A poor grade does not prove provider fault, especially on Wi-Fi. Compare controlled runs, retain the actual idle and loaded numbers and use application-specific latency where available.
If the rise appears only on Wi-Fi, improve placement, interference or wireless backhaul before changing the broadband package. If it remains over Ethernet, check whether the router offers documented Smart Queue Management, SQM, CAKE, FQ-CoDel or an equivalent queue-control feature. Save the original configuration before changing anything.
Queue management often requires a configured rate below the real line capacity and can reduce maximum benchmark speed in exchange for responsiveness. Avoid copying unexplained settings from forums. If the router cannot manage the connection rate, seek model-specific guidance or provider support before purchasing replacement hardware.
Read the complete LumaProbe test methodology and privacy notes →
It uses the larger increase from idle latency to download-loaded or upload-loaded latency: A up to 15 ms, B up to 30, C up to 60, D up to 100 and E above 100.
No. Queueing can occur in the device, Wi-Fi, router or provider path. Repeat over Ethernet and under controlled conditions before assigning the cause.
Cloud backups, file sharing and camera uploads can fill an upstream queue. A download-only test can miss that source of lag.
Some links may earn us a commission at no extra cost to you. Recommendations are categories, not paid rankings.
Affiliate disclosure