Starts only when you press Start test. Downloads and uploads may use hundreds of MB or more on a fast connection. Cloudflare receives your IP address and test traffic. If you click the optional ISP lookup, ipwho.is also receives your IP address. Prefer an unmetered connection.
How to get a useful speed-test result
- Choose the connection you want to check. For a baseline, connect with Ethernet; to investigate Wi-Fi, test in the room where you normally use it.
- Pause large downloads, cloud backups and streams on other devices. Keep this tab visible. Note whether you are using a VPN so that later comparisons use the same setup.
- Press Start test once. The button becomes Stop test while the measurement runs. Wait for the completed result before comparing it with another run.
- Repeat two or three times at different times of day. Compare the typical result and the spread, rather than treating the fastest run as your everyday speed.
Many tests finish in about 20–30 seconds, but high latency or network errors can take longer. A stopped or failed run can leave partial numbers on screen; those are not saved as a completed result. The test stops if you leave the page, and a 70-second overall timeout prevents an indefinitely running test.
How to read download, upload, ping and jitter
- Download measures incoming data in Mbps. It affects large file downloads and how much streaming traffic your connection can carry.
- Upload measures outgoing data. It matters when sharing files, backing up photos, broadcasting video or sending your camera feed in a call.
- Ping is the measured idle round-trip time in milliseconds. Lower usually means a more responsive connection to this test endpoint.
- Jitter describes changes between consecutive ping samples. A low typical ping with irregular spikes can still make calls and games uncomfortable.
- Ping under load is measured during the download phase. A large increase can indicate queueing somewhere along the path; this result alone cannot identify which device causes it.
- Estimated data totals the application bytes counted during download and upload, including warm-up. It is not a mobile-provider billing counter: protocol overhead, the small ping requests and some browser buffering are not included.
Worked example: high speed with slower responsiveness
Suppose the completed tiles show 200 Mbps download, 20 Mbps upload, 18 ms ping, 3 ms jitter and 95 ms ping under load. These are illustrative numbers, not a measurement of your connection. The download rate is about 25 MB/s, because 200 ÷ 8 = 25. The loaded ping rises by 77 ms, because 95 − 18 = 77.
For this example, all four activity cards meet this tool's higher thresholds. The summary still warns that latency increased during download. Those statements answer different questions: the connection has enough measured capacity for the activity guide, while busy traffic may make it less responsive. A game can also use a different, more distant server.
At a constant 25 MB/s, a 1 GB decimal download would take about 40 seconds before overhead: 1,000 MB ÷ 25 MB/s. Actual transfers can take longer because of the remote server, Wi-Fi, competing traffic and the time needed to reach full speed.
How the activity cards are decided
These are this site's practical rules of thumb, not universal service requirements. They evaluate only this completed test. A real app may need more or less bandwidth, and several people share the same connection.
The higher result takes priority; anything below the middle conditions receives the lower result. If no loaded-ping sample is available, the gaming card uses idle ping and jitter without that extra condition. It does not establish that loaded latency is good. These cards do not check packet loss, call software, video resolution settings or the route to a game server.
For comparison, Netflix's own speed recommendations currently list 3 Mbps for 720p, 5 Mbps for 1080p and 15 Mbps for 4K. Our 25 Mbps 4K threshold is deliberately more conservative. Check the service you actually use and leave room for other household activity.
What to try when the speed looks low
- Separate Wi-Fi from internet access. Compare a wired run with a run near the router and one in your usual room. A large difference suggests the local wireless setup deserves attention.
- Remove competing traffic. Pause game updates, cloud sync and streaming, including on other devices. Then repeat under the same conditions.
- Compare VPN configurations if appropriate. A VPN changes the route and can affect speed. Test with the configuration you normally need; do not disable one required by your workplace or network policy.
- Keep the browser active. Background tabs, battery-saving modes and a heavily loaded device can affect a browser measurement. Try another current browser or device if the result is unexpectedly different.
- Investigate busy-connection lag. If loaded ping rises while idle ping stays low, reduce background uploads/downloads or investigate your router's queue-management settings. The test measures loaded latency during download only.
- Record a repeatable problem. If several wired tests remain far below expectations, compare with another independent test and contact your provider with times, connection type and completed results.
If the test will not finish
Check whether your browser, VPN, firewall or network policy permits requests to speed.cloudflare.com. Do not disable protections you rely on just to run a test; use an approved network or retry later. A third-party outage can also prevent completion. The optional ISP-name lookup is separate: a failed lookup does not mean your connection has no ISP or that a completed speed result is invalid.
Measurement method and limits
This page uses its own browser measurement code with Cloudflare's documented download and upload endpoints; it does not run Cloudflare's complete speed-test application. Two initial latency requests warm up the connection and are discarded. The next 12 requests supply the median ping; jitter is the mean absolute difference between consecutive samples.
Download and upload each run for approximately eight seconds, with up to six simultaneous request streams. Browsers may share transport connections. The first 1.5 seconds of each transfer phase are excluded from the reported average speed. Request sizes adapt, up to 100 million bytes per download request and 16 million bytes per upload request. Upload payloads are generated test bytes, not your files.
After the first 1.5 seconds of download, additional latency requests run with a pause between them. Their median becomes ping under load. The tool does not measure packet loss, upload-loaded latency, Wi-Fi signal strength or performance to every website. Browser timing, device limits, protocol overhead, server availability and competing traffic all affect the estimate.
Privacy, test traffic and saved history
Network measurements contact Cloudflare only when you start the test. Cloudflare receives your IP address and test traffic. Clicking Look up ISP name separately contacts ipwho.is, which also receives your IP address. The IP display is hidden until you choose Show; hiding it on screen does not hide it from these services.
The tool does not post results to Your Tech Tool. Up to five completed results stay in this browser's local storage under ytt_speed: time, download, upload, ping and jitter. History does not retain your IP, loaded ping or transferred-data total. Clear history removes those saved entries. Copy result copies download, upload, ping and jitter from the current completed run.
Fast connections can use hundreds of megabytes or more. For scale, eight seconds at a steady 200 Mbps down plus eight seconds at 20 Mbps up transfers about 220 MB: (200 + 20) × 8 ÷ 8. This is a hypothetical constant-rate calculation, not a prediction from the final speed tiles. Use an unmetered connection if you are concerned about data charges.
Speed test FAQ
Is Mbps the same as MB/s?
No. Mbps means megabits per second; MB/s means megabytes per second. Divide Mbps by eight for a decimal MB/s estimate. File-size displays using MiB use a different, binary unit, so their numbers will not match that conversion exactly.
Why is upload slower than download?
Some internet plans provide different capacities in each direction. Wi-Fi, VPNs, device limits and the path to the endpoint can also affect the two phases differently. Compare the result with the upload allowance in your particular plan.
Why do other speed tests show different numbers?
Tests can use different servers, request patterns, durations and averaging rules. This one measures browser traffic to Cloudflare. Compare repeated tests in similar conditions rather than expecting every service to return the same value.
Can this prove that my provider is at fault?
No. A single browser test cannot isolate the provider from Wi-Fi, your device, a VPN, congestion or the remote service. Use wired comparisons and repeated measurements to assemble useful evidence.