Why Do Speed Tests Give Different Results?

Run three different speed tests one after the other and you'll often get three different numbers. That doesn't mean two of them are broken. A speed test doesn't measure "your internet" in the abstract — it measures one path, to one server, with one method, at one moment.

The main reasons

• Different servers. A server in your city and one in another country give different speeds and ping. Some tests pick the closest server; others pick one with spare capacity.
• Inside or outside your provider's network. Some servers sit inside your provider's own network ("on-net"), which avoids the busy links between networks. Others sit outside ("off-net"), closer to how you reach real websites.
• Single vs. multiple connections. Some tests use one connection, others several in parallel. Several connections often show a higher speed, especially on fast or long-distance links, because they fill the line more easily.
• Warm-up handling. Connections start slowly and speed up. Tests that leave out the start show higher averages than tests that include it.
• Test length. A short test may finish before the speed settles; a longer one evens out bursts.
• Timing. Your network and your provider's change constantly — evenings are busier.
• Your device and Wi-Fi on that particular run, including other devices using the connection.

Why ping differs too

Ping depends mostly on distance to the server, so a test using a server 20 km away will show lower ping than one using a server 500 km away. Tests also measure ping in different ways — before the download starts, or during it — and report the lowest, average or middle value. Our test reports idle ping and also "loaded" ping during download and upload, which shows bufferbloat.

How our test works

We use M-Lab servers by default (one connection per direction) and our own server as a fallback (several connections), leave out the first second, and show which server measured each result. M-Lab's servers are off-net — outside your provider's network — which reflects the path to most websites rather than your provider's best case. Full details are in how it works.

Which number is "right"?

All of them, for what they measure. A higher number from an on-net, multi-connection test shows what your line can do in ideal conditions. A lower number from an off-net, single-connection test can be closer to what one download from a distant website gets. Neither is a lie.

What matters most is consistency. Pick one test, use it the same way each time, and watch for changes. A drop from your usual result tells you far more than a comparison between two different tests.

How to compare fairly

1. Use a cable and the same device. See Wi-Fi vs Ethernet.
2. Pause downloads, backups and streams on other devices.
3. Run each test two or three times, back to back.
4. Look at typical results, not the best one.
5. Test at the same time of day, and note which server each test used.

For a step-by-step guide, read how to test your internet speed.

When different results point to a real problem

If every test, on a cable, at different times, shows far less than your plan, the problem is real — not the test. Keep a few results with dates and times, then work through why is my internet slow before contacting your provider. If only certain services are slow, read is my ISP throttling my internet.

Browser and device effects

On very fast connections, the device itself can become the limit: an older phone, a laptop on battery saver, or a browser with many tabs open may not keep up with gigabit speeds. Close other tabs, plug laptops in, and use the same browser each time you compare.

Questions