YOUR SPEED. YOUR DEVICE.
Why Internet Speed Tests Give Different Results
Internet speed tests give different results because they measure different things: which server they use, how many connections they open, how long they run and when you run them. M-Lab's NDT, which Xpeed uses, sends data over a single connection to a server outside your provider's network. A test that opens several connections to a nearby server can report a higher number on the same line at the same time.
XpeedOne connection or several
M-Lab describes NDT as a single-stream measurement of a connection's capacity for bulk transport. Its FAQ says the test tries to transfer as much data as it can in ten seconds in each direction, using a single connection to an M-Lab server. Other popular tests transfer data across multiple connections at once.
Several parallel connections can fill a link that one connection cannot, especially when some packets are lost along the way. A single connection is more sensitive to loss and delay on the path, which is why M-Lab considers it more useful for finding network problems. The single-stream figure shows what one transfer achieves; a multi-stream figure shows what several achieve together.
The ndt7 version uses the BBR congestion control method where available and falls back to Cubic otherwise, according to M-Lab. Tests that use other methods can reach a different rate on the same path.
Where the test server is
M-Lab states that its measurements are always conducted off-net. Your data crosses from your provider's network into other networks to reach the M-Lab server.
Some tests use servers inside or close to your provider's network. M-Lab notes that such on-net results are often higher, because the path is shorter and one provider controls all of it. An off-net result includes the links between networks that your traffic also crosses when you browse or stream.
When you run the test
M-Lab's FAQ notes that a test run five minutes ago can show very different results from one run twenty minutes ago. Routes change, and so does the amount of other traffic on each part of the path.
A test at noon and another late in the evening run under two different sets of conditions, at home and along the route. Compare them as such.
Your own network
Many slow results start inside the home. The FCC notes that several wireless devices on one Wi-Fi network can reduce performance. It describes 2.4 GHz Wi-Fi as having broader coverage but processing data less quickly than 5 GHz, which is faster over a shorter range and may be less crowded.
According to the FCC, a direct ethernet cable between the router and the device gives the highest speeds and avoids Wi-Fi congestion. The FCC also suggests restarting the router and checking whether an outdated router limits the speed you pay for.
A VPN sends your traffic through another server first, which changes the path the test measures. Downloads, video streams and backups running on other devices take a share of the same connection.
How to compare results fairly
Keep everything the same except the thing you want to compare:
- Use the same device, browser and connection type, preferably a cable to the router.
- Pause downloads, streaming and backups on the network.
- Turn the VPN off unless the VPN is what you want to measure.
- Run each test several times at different times of day.
- Write down the server, time and conditions with every result.
When two different tests still disagree, look at their methods before their numbers. A multi-connection test to a nearby server and a single-connection test to an off-net server measure two different workloads.
What one result does not prove
A single run describes one path at one moment. It does not show what you will get from every website, game or video service, and it does not prove that a provider is limiting a particular test. A pattern repeated across many runs and conditions tells you more than any one number.
Each Xpeed run sends and receives real data, which counts toward a metered plan. M-Lab receives and publishes the test data, including your IP address, under its own privacy policy. Xpeed asks for consent before every test.
Questions and answers
Which speed test is correct?
Each one is correct for what it measures. A single-connection test to an off-net server and a multi-connection test to a nearby server answer different questions, so their results can differ on the same line.
Why is my Wi-Fi result lower than my plan's speed?
Wi-Fi performance drops when many devices share the network, with distance from the router and with crowded frequency bands. The FCC recommends a wired ethernet connection for the highest speeds.
How long does the Xpeed test take?
M-Lab's NDT tries to transfer as much data as it can in ten seconds for download and ten seconds for upload. Xpeed runs download first, then upload.
Does a low result mean my provider is throttling?
Not by itself. Server location, the number of connections, home Wi-Fi and the time of day all change the result. Repeat the test under the same conditions before drawing conclusions.