Internet Speed Test

Check whether the connection is metered before starting. Close large transfers you do not want competing with this measurement, or deliberately leave normal activity running if your question is about the connection under its usual load. Record which condition you chose so later comparisons mean the same thing.

Read the transfer allowance and select the consent checkbox. Start the internet speed test only when you are ready for real network traffic. The bounded sequence includes small and larger download and upload samples, with optional result-reporting endpoints disabled. Cloudflare still receives the traffic and your request address.

Wait for completion or use Stop test. An interrupted sequence is not presented as a final result. If some measurements fail, available values remain labeled as partial and missing values show unavailable rather than zero. Do not treat zero and missing as equivalent when reporting a connection problem.

Compare download and upload separately. They describe different directions and may differ substantially on the same connection. Read Mbps as megabits per second, not megabytes per second. Dividing a Mbps value by eight gives a rough bytes-per-second conversion before protocol overhead and application behavior.

An internet speed test estimates throughput across a particular route during its measurement window. The browser, device, network activity and remote endpoint all participate in that result. This component uses Cloudflare’s published engine with a limited transfer sequence so the data allowance is explicit. A short, bounded test may underrepresent a very fast connection because large transfers can be needed to reach steady throughput. It is useful for a quick repeatable observation, not a replacement for an ISP contract measurement procedure or a comprehensive network assessment.

Wi-Fi speed test searches often refer to internet throughput while connected by Wi-Fi. This tool does not measure radio signal strength, negotiated wireless link rate or the isolated speed between your laptop and router.

A VPN can change the route and add overhead, while a busy device can limit browser performance. Keep these factors visible when comparing results. A single low observation does not establish which component caused the limitation.

The internet speed test does not collect a WebRTC packet-loss sample, run a hidden background transfer or automatically save a monitoring history. Repeat only when a new observation will answer a specific comparison question.

A hypothetical result of 80 Mbps download corresponds to about 10 megabytes per second before overhead, because one byte contains eight bits. It does not mean that every file download will sustain that rate. A small file can spend a larger share of its transfer on connection setup, while a busy server can deliver more slowly. Upload may have a different limit altogether. When comparing two internet speed test results, preserve the endpoint, transfer method and connection state instead of comparing only the largest number displayed.

Keep the question, input, observation time and source together. A bare address, percentage or status code is difficult to interpret later because it omits the connection and measurement context. When comparing two observations, note what changed between them: the target, network, VPN state, selected adapter, record type or remote origin. Change one relevant condition at a time when possible. That makes the comparison easier to explain without inventing a cause from one number.

Save only information you actually need. Network records can reveal infrastructure names and connection details, even when they do not contain a password. Remove unrelated identifiers before sharing a screenshot with another person. Never put private credentials in a domain field or publish an internal hostname as part of a public test. A copied result is a snapshot, not a promise that the same conditions still apply when someone reads it later.

Kepo is a Mac desktop-widget product for repeated checks and small workflows. A labeled troubleshooting note or a supported monitoring widget can keep the relevant context accessible while you work. The download link opens the product workflow; this article does not automatically install a configured diagnostic widget, transfer your network inputs or start a scheduled check. Confirm the actual widget’s capabilities before relying on it for ongoing observations.