RunTheTests
/

Download Speed Test

Download speed, and what it actually supports — which is the question behind it.

Run the test

Measured against one Cloudflare edge with 4 parallel streams. A commercial speed test picks a server from a large fleet, often on your own provider's network — this measures the path to a CDN, which is closer to what a real download does and reads lower than the figure your provider quotes.

How this check works

Four parallel streams pull one-megabyte chunks for eight seconds and the steady-state rate is taken from what is left after the ramp is discarded. That discard is the part that matters: a TCP connection does not start at full speed, it probes upward, so averaging across the whole transfer reports a fraction of a fast line — which is why so many browser-based speed tests read low. The result leads with what the speed supports rather than the number, because the figures involved are startlingly small next to what people are sold.

What the results mean

Ramp discarded
The chunks dropped as slow-start. Without discarding them, a fast connection measures as a slow one.
What it supports
The published requirements for each activity. A video call wants around 4 Mbps and a 4K stream around 25 — far less than most packages provide.
One edge
The measurement is against a single CDN edge rather than a chosen nearby server, which is closer to a real download and reads lower than a provider's quoted figure.
Bandwidth is not the usual problem
Calls and games break on loss and jitter, not on capacity. A high number here alongside a poor call is the normal shape of that.

Common problems and fixes

The result is far below what I pay for
Expected to some degree — a browser tab against one CDN edge is not the same measurement as a provider's own test. Compare over a cable and with nothing else running before concluding anything.
The speed is fine and video calls still break up
That is the common case, and bandwidth is not the cause. Run the bufferbloat and call-readiness tests: loss and queueing ruin calls at any speed.
Wi-Fi is much slower than the cable
Normal, and worth quantifying. The difference is your wireless network rather than your connection, and it is the number to act on.

Frequently asked questions

Why is my speed test result lower than my package?

Several reasons compound. This measures against one CDN edge rather than a server chosen for proximity, a browser tab has overheads a native client does not, and Wi-Fi typically costs a large share of a fast connection. The provider's quoted figure is also usually a maximum to the exchange rather than a promise to any particular destination.

How much speed do I actually need?

Less than most people are sold. A video call wants around 4 Mbps, HD streaming around 8, and 4K around 25 — so a 50 Mbps connection comfortably carries a household doing all three at once. What tends to break is not capacity: it is packets arriving late or not at all, which is what the bufferbloat and stability tools here measure.

Does this test upload as well?

Separately — the upload speed test is its own page. Upload is worth measuring in its own right: home connections have far less of it, and it is what limits video calls and backups.

More in Connection