What your speed-test results mean
Download measures data received by your browser; upload measures data sent from it. Both are rates. Mbps uses megabits per second and MB/s uses megabytes per second: divide Mbps by eight to convert between them.
For example, 100 Mbps equals 12.5 MB/s. At a constant 100 Mbps, a decimal 1 GB file (1,000 MB) takes 1,000 ÷ 12.5 = 80 seconds before extra overhead. This is an ideal estimate, not a promise about a particular download service.
| Rate (Mbps) | Rate (MB/s) | 100 MB | 1 GB |
|---|---|---|---|
| 10 | 1.25 | 80 seconds | 800 seconds |
| 25 | 3.125 | 32 seconds | 320 seconds |
| 50 | 6.25 | 16 seconds | 160 seconds |
| 100 | 12.5 | 8 seconds | 80 seconds |
| 200 | 25 | 4 seconds | 40 seconds |
File applications may use binary units such as MiB or GiB. Those differ from the decimal MB and GB used in this table.
How this browser test works
The test sends ten small HTTPS latency probes to Cloudflare, then downloads up to 51 MB and uploads up to 20 MB of synthetic payloads in sequential requests. A transfer phase stops requesting larger payloads after a completed request takes the phase past 4.5 seconds. Each request is limited to 15 seconds, and a run is limited to 60 seconds.
Download and upload are payload bits divided by elapsed time in decimal megabits per second. Request setup, browser scheduling and response time are included. Small capped transfers can understate fast connections; results are estimates for this device and route, not a certification of an ISP plan.
Latency is the median successful HTTPS request timing. Jitter is the mean absolute difference between consecutive successful timings and is unavailable with fewer than two samples. These measurements do not measure ICMP ping, transport packet loss, loaded latency or a particular game server.
This implementation uses Cloudflare endpoints with its own short measurement sequence. It does not reproduce all measurements or scoring used by Cloudflare’s official test, and matching another speed-test service is not guaranteed.
Cloudflare’s measurement engine documentationCompare results without changing several things at once
- Keep the device and browser the same. Pause downloads and backups, and keep this tab active.
- Record whether you are using Wi-Fi, Ethernet, mobile data or a VPN. Test the connection you actually intend to use.
- Repeat at another time, or change one condition such as moving nearer to the router. Compare the repeated pattern.
- Check the actual service when troubleshooting calls or games. This endpoint can follow a different network route.
Use the continuous ping test to investigate HTTPS timing over a longer interval, or the WebRTC leak test to investigate possible public-IP exposure. Neither replaces a throughput measurement.
Internet speed test questions
How much data does the speed test use?
One run can download up to 51 MB and upload up to 20 MB of synthetic data, plus protocol and lookup overhead. A slow connection, cancellation or request timeout may end the test earlier. Repeated tests consume more data.
Why does the result differ from my ISP plan or another speed test?
Tests can use different servers, routes, durations, parallel connections and timing methods. This test uses capped sequential HTTPS transfers. Wi-Fi conditions, device load, VPN routing and other traffic also affect the result. Compare repeated runs using the same setup.
What is the difference between Mbps and MB/s?
Mbps means megabits per second; MB/s means megabytes per second. Both are transfer rates. Divide decimal Mbps by eight to get decimal MB/s: 100 Mbps equals 12.5 MB/s. Actual file completion time also includes overhead and possible changes in throughput.
Does this test prove that gaming or video calls will work well?
No. The measurements describe this browser’s route to the test endpoint. Games and calls may use different servers and protocols. Upload capacity, packet loss, loaded latency and competing traffic can affect them even when download speed is high.
Which services receive connection information?
Starting the test contacts Cloudflare for measurement traffic and Cloudflare/ipapi for IP and provider context. Those services receive your connection IP and use their own policies. The tool does not select personal files, and results remain in page memory unless you copy them. Provider logging cannot be ruled out.
How should I compare Wi-Fi and Ethernet?
Use the same device, test endpoint and VPN setting. Pause background transfers and keep the tab active. Run several tests on each connection and compare the pattern; one result cannot isolate the cause of a slowdown.
Do I need an internet connection, and are my inputs uploaded?
An internet connection is required. The site handshake sends a random challenge, not files or text inputs. Network diagnostics need external connections. IP checks contact Cloudflare and ipapi; DNS queries send the entered hostname to Cloudflare; WebRTC contacts ipify for IPv4 and IPv6 detection and Google STUN; latency probes contact the listed providers. Those services receive your connection address and may apply their own logging policies. Results are not saved by this tool.