Internet Speed Test

How this internet speed test works

This internet speed test measures the route between your browser and the selected server in three phases. It discards one warm-up request, then sends 12 small no-store requests. The median response is ping; the mean absolute difference between successive responses is jitter. The smallest response appears as a secondary detail.

For download, four parallel streams loop over a cache-busted, incompressible payload. The first 1.8 seconds count as ramp-up, keeping connection setup and slow-start from dominating the final rate. The test reports bytes received during the steady window, for up to about 10 seconds or 60 MB. If the measured rate is still climbing at three seconds, it can add two streams.

Upload uses three parallel XMLHttpRequest streams. Each request starts at 1 MB and can grow to 4 MB if it finishes quickly. Progress events count bytes as they are sent, and the server discards the random body. If the PHP endpoint is unavailable, the page reports that instead of making up an upload number.

During both transfers, a small ping runs about every 400 ms. The result is loaded latency. It shows the worst phase's median latency first, then download and upload values with their increases over idle. The A to D label grades the worst increase: A is under 30 ms extra, B under 60 ms, C under 200 ms, and D is 200 ms or more.

Download, upload, ping and jitter explained

• Download (Mbps) is how quickly data arrives. It matters for streaming, browsing, updates and downloads.
• Upload (Mbps) is how quickly data leaves. It matters for camera calls, live streaming, cloud backup and large file transfers. If upload is the weak direction, test over Ethernet before blaming the plan. Many home plans are deliberately asymmetric.
• Ping (ms) is the round-trip time to this server. Lower is more responsive, but another service or game may use a different route.
• Jitter (ms) is variation between the idle readings. A connection can have ample bandwidth and still sound or feel uneven when jitter is high.
• Loaded latency shows whether queues add delay while the connection is busy. A large increase points to bufferbloat or competing traffic, but does not automatically mean the access plan is too slow.
• Mbps and MB/s describe different units. The toggle converts the throughput display: 8 megabits (Mb) make 1 megabyte (MB), so 100 Mbps is about 12.5 MB/s.

This page does not measure packet loss. If calls or games stutter while the speed looks good, read what packet loss means and test the destination you care about.

What is a good internet speed?

There is no universal cutoff. A good result depends on which activities overlap at home and on route quality. Netflix lists 15 Mbps for one 4K stream, Zoom publishes direction-specific call requirements, and the FCC's 100/20 Mbps figure is a broadband benchmark, not a personal minimum. The good internet speed guide gives practical activity examples. The packet loss guide explains why dropped data matters.

Why results differ between test sites

Each tester measures a specific route, server, device, and moment. A multi-server test may choose a nearby host, while this page can use Cloudflare automatically or the Loqmi server in Germany. Stream counts, payload sizes, ramp-up rules, browser scheduling, and server capacity can also affect the result.

For a useful comparison, pause large transfers, use the same device and location, and repeat under the same conditions. Compare Wi-Fi with Ethernet when possible. This result describes this host. It does not predict performance on a game server, video provider, or in every room at home.

How to troubleshoot a slow connection

• Run the test again with cloud backup, updates, streams and game downloads paused. Active traffic can affect both throughput and loaded latency.
• Compare the problem location with a test close to the router or over Ethernet. If Ethernet is strong but one room is weak, the Wi-Fi in that room is the bottleneck: move closer, wire the device or reposition the router before changing plans.
• Check both directions. A download-heavy plan does not solve an upload bottleneck during calls, live broadcasts or backup.
• Look for a pattern across repeated tests rather than treating one reading as a permanent line speed. Keep the conditions with the result when contacting your provider.
• If latency rises sharply under load, reduce simultaneous transfers and investigate queue management on the router. A faster tier may help capacity, but it cannot repair every route or wireless problem.

Cloudflare may select a nearby data center, shown by its IATA code when available. If you choose the Loqmi server, it is in Germany. Distance and routing matter when you compare this result with another service.

Frequently asked questions

What does this test measure?

It counts real bytes transferred between your browser and this server, uses parallel streams for throughput, discards a warm-up window, and measures idle and loaded latency separately. It is a useful sample of this route, not a guarantee for every service or game server.

What are ping, jitter and loaded latency?

Ping is the median idle round-trip time. Jitter is the mean absolute difference between successive idle readings. Loaded latency shows the worse of the download and upload medians, while the grade uses the larger increase over idle.

How much speed do I need?

Use the activity that matters rather than a universal number. The linked internet speed guide uses published service requirements and explains why simultaneous use changes the answer.

Why do results differ between speed test sites?

Servers, routes, stream counts, device conditions and timing differ. Repeat with a consistent setup, and compare wired and Wi-Fi results when diagnosing a home-network problem.

Does this speed test store my data?

With the Loqmi server, random test payloads are discarded after counting. When Cloudflare is selected, the test exchanges data with Cloudflare, which may record the measurement. The last five result summaries are stored only in this browser, are never sent to the server, and can be cleared below the tool. IP lookup results are not saved in browser storage; the lookup endpoint retains the request address only in standard hosting logs. See the privacy policy.