How good is my connection?

This page measures what reaches this device, right now, on this network. It does not measure your line: if you are on Wi-Fi at the far end of the corridor, it measures the corridor. That distinction matters, and I come back to it below.

How the measurement works

Response time is measured by asking twelve times for a two-byte file and looking at how long passes between request and reply. This is not the ping of video games, which uses a different protocol: it is the round trip of a web request to the nearest server of the network hosting this site. For working out whether a line is fit for gaming it does the job; for comparing with the figure your console shows, it does not.

Stability is the average variation between one request and the next. It is the number that explains video calls seizing up while the speed test says everything is fine: for talking to people, a fast but irregular line is worse than a slow steady one.

Download is measured by fetching the same block of random data four times in parallel, discarding the first half second, which is where the connection is still picking up speed. The test stops at whichever limit comes first: four and a half seconds, or twenty-four megabytes downloaded. The cap is deliberate, for two reasons: a test that does not stop moves hundreds of megabytes per person, and this site has no network to give away; and above a certain speed the bottleneck becomes the browser rather than the line.

A practical consequence worth knowing: on a very fast connection this measurement underestimates, because the test ends before the transfer reaches full speed. Where you start to notice is not a fixed number, since it also depends on the browser, the device and the network path: on the commonest Italian fibre lines it happens somewhere above 300 or 400 Mbit/s, but treat that as an order of magnitude. Either way, with fast fibre read the result as a floor rather than a ceiling.

What this page cannot measure, and why

There is no Wi-Fi analyser in here, and not out of laziness: a web page has no access to the name of the network you are on, to signal strength, to the channel or to the band. An installed app with the right permissions sees those things, and the browser does not expose them to any site. What the browser does offer is a coarse estimate of the connection type, called the Network Information API, which exists on Chrome and does not exist at all on Firefox and Safari: it is an estimate, declared as such, rather than a measurement, and I preferred not to show it rather than pass a label off as data.

Nor is there a scan of the devices in your home. Since 2025 browsers have been closing that road for security reasons: reaching a local network address from a web page requires an explicit permission from the user, and you have to know the address already. A page that walks the network to see who answers is exactly what that permission exists to prevent, and rightly so. If I ever make an installable version, that one will be able to do it.

Upload is measured now, and it works the other way round from download: the browser sends real data to an address on this site that receives it and throws it away, storing nothing anywhere. There is a cap here too, a lower one: four seconds or twelve megabytes. It is lower because on almost every home line the way up is far slower than the way down, and with the same cap as download the test would take forever. If the upload box stays empty it means not a single block made it through: that happens on a very slow network, or when a corporate filter blocks the sending, and I would rather show nothing than a made-up number.

One thing worth knowing before you press the button: between down and up, a full test moves a little under forty megabytes. On a home network you will not notice, on a mobile connection with a few gigabytes a month you will. The exact figure for each test is in the "data used for the test" box, which is there for that reason.

If the numbers are bad

Before blaming the provider it pays to rule out the things that depend on your home, which is most of them. The simplest test is to repeat this measurement with the computer plugged into the router by cable, if you can: if the numbers improve a lot, the problem is the Wi-Fi and not the line. If you cannot, repeat it standing next to the router and then in the room where you usually complain.

The checks in the right order, from where the router sits to crowded channels and what to ask whoever sells you the connection, are in The internet is slow: the checks to make before calling support.

The measurements you take will appear here, most recent first. They stay on this device and are not sent anywhere.