Why Do Internet Speed Tests Give Different Results?
I am sorry. Last week in Tech 911 , Lifehacker’s tech tips column, I answered a question from Lifehacker editor-in-chief Melissa . She technically reads Lifehacker, so it still fits the theme of this column, but looks like a cheat. I’m doing the same thing again this week, not because you didn’t send in valuable questions – yes! – just another Lifehacker employee had a great question on a topic we recently researched and I’m sure some of you will have similar thoughts.
Lifehacker social media editor Tim writes (via Slack):
“Why the hell do internet speed tests work and why am I getting completely different results on different tests?!?!”
I researched this question a little more, and it turned out that Tim was seeing quite a few different numbers when running various web speed tests on his home connection. On one? 178.4 Mbps isn’t too bad. With another? 187. Still great. However, on the third, we are faced with a strange problem: 22.7 Mbps, which looks like a disaster. It can also be a red herring, launching a person into a tornado of troubleshooting they may not even need to.
First, all speed testing sites calculate statistics a little differently, such as download / upload speed, but the general principles are pretty similar. For example, this is how Speedtest.net describes its process .
To measure latency, a site checks the time it takes for a message to bounce from you, to the server, back to you, and repeats the process over and over — the lowest value is the result of latency. For uploads and downloads, Speedtest measures how long it takes to process multiple chunks of data while simultaneously filling the feed with data (sort of) throughout the test.
Here’s Netflix ‘s approach to fast.com :
“Based on network bandwidth, the fast.com client runs the test using a variable number of concurrent connections. For low bandwidth networks, launching more connections can cause each connection to compete for a very limited bandwidth, resulting in increased latency and a longer, less accurate test.
However, when the bandwidth is high enough, launching more concurrent connections helps to saturate the network faster and shorten testing time. For very high bandwidth connections, especially in higher latency situations, a single connection and a 25MB file may not be enough to achieve maximum speed, so multiple connections are required.
For each connection, the fast.com client chooses the size of the 25MB file chunk it wants to download. In situations where the network layer supports periodic execution events, it makes sense to query the entire file and estimate the network speed using download progress counters. In cases where the download progress event is not available, the client will gradually increase the payload size during the test in order to perform multiple downloads and get enough samples. “
Even if this sounds like a bunch of network nonsense to you, it’s okay. The point is that you have to approach bandwidth testing scientifically because there are many variables that can affect a typical Internet speed test, that is, in addition to the inherent differences.
Perhaps you have an artificially slow connection to some of the server (s) used by the test, or the test is failing. Your ISP may prioritize or limit your performance depending on how the test is going. Even your browser full of add-ons and extensions can interfere with a certain test in some way. Or you run tests at different times and misinterpret a “bad” internet connection for what is actually due to network congestion from anyone in your area streaming Netflix at the same time – that sort of thing.
And don’t make me run these tests on a wireless device. Use a wired connection whenever possible to exclude variables (if you’re only trying to figure out if your house or apartment is actually getting the speed you pay for). Adding a wireless network makes troubleshooting much more difficult as you have to run more tests to pinpoint the problem with your ISP, router, or device.
Treat internet speed tests like a science project
You don’t want to rely on the results of one test from one service, even if they seem great. You should run some tests from multiple vendors to see if you can understand the trend . Is your speed on average 75 percent slower than you pay? There is a problem.
One or two tests show average speed, and all other sites are okay? You are probably fine, especially if you haven’t noticed any major problems with day to day use.
Are you having trouble uploading gigabit files over yoru fiber optic connection? Perhaps your laptop – with its mechanical hard drive and small amount of memory – cannot handle a huge amount of data, and this affects the test results (to make the process much easier ).
When I want to get a good idea of my internet speed, I do various tests and see how things are going before I get stressed. This list includes:
Sounds like a lot, but all of these tests are easy to run – you press a button, you puff on it, and you go to do another one. It’s simple.
If one or two tests fail, it doesn’t really matter. If I can’t get the speed I expect on any of them, I think how much I want to ruin my network setup at home (or my desktop) because I don’t get the full 100+ Mbps that my roommates and I are should have fun. Fun times!