Wirebook › Bonus lesson
Why does every speed test give a different answer?
Run three speed tests and you'll get three different numbers. One site says 140, another says 60, and the download on your games console says 17.8. Is somebody lying? Not really. A speed test is like a water-balloon race: the site fills a balloon from the internet tap for about ten seconds, then divides the water by the time. How fast a balloon fills depends on much more than the size of your pipe.
Why does a hose start slowly?
The computer sending the water has no idea how wide your pipe is. If it blasted everything at once it would flood the pipes and lose most of it. So every connection starts with a small squirt and waits to hear back that it arrived. If it did, it sends twice as much. Then twice as much again. Engineers call this slow start, even though the doubling soon makes it quick.
The catch is that each doubling has to wait for the reply to come back: the ping. A tap down the road answers in about 8 thousandths of a second, so the hose opens up almost instantly. A tap on the other side of the world takes around 220 thousandths, nearly a quarter of a second, for every doubling. On a ten-second test, that slow opening drags the average down. That's why speed test sites try to pick a server near you, and why a few let you leave out the first couple of seconds.
What does a leaky hose do?
Data travels in small packets, like drops of water, and now and then one goes missing: a crackly line, a busy Wi-Fi channel, a long trip through lots of equipment. The sender can't tell a lost drop from an overflowing pipe, so it plays safe and eases back to about 70% of its speed, then creeps up again. A hose that keeps losing drops never gets to open up fully. That's packet loss, and it hurts most on far-away taps, because creeping back up takes one round trip per step.
Here's the clever trick: use several hoses at once. When one hose loses a drop, only that hose eases back; the others keep pouring. Most speed tests open several connections together, which is why they can read higher than a single big download from one server. Neither number is wrong. They're measuring different races.
Why does my download say a much smaller number?
Often it isn't smaller at all. Internet plans and speed tests count in megabits per second (Mb/s). Download apps, game launchers and browsers usually count in megabytes per second (MB/s). A byte is 8 bits, so divide by 8: a 150 Mb/s plan that's working perfectly downloads at about 17 or 18 MB/s. The little b and the big B make all the difference.
And plans are measured before the packaging. Every packet carries an address label and some checking information, so about 5% of the water is the bucket, not the water. A test showing around 94% of your plan on a cable means everything's working.
So why do different sites disagree?
- Different taps. Each site uses its own servers in its own places. Some are next door; some, especially when a site is busy, are far away.
- Different numbers of hoses. Some open one connection, some open eight or more. One hose is the honest picture of a single download; many hoses show what the line can do.
- Different rules. Some leave out the slow start or the slowest moments, some test for longer, some stop early.
- Different moments. Run two tests a minute apart and your neighbours' films, the Wi-Fi, and whatever else is happening at home will all have changed.
- Your provider's own test often runs to a server inside its network, so it skips the busy roads between networks. It shows your line, not the whole internet.
The way to get a fair answer: plug in a cable, pause downloads and updates, pick a server near you, test a few times at different times of day, and compare like with like.
So why does every speed test give a different answer?
Because a speed test doesn't measure your pipe directly. It times how fast one balloon fills, from one tap, through a few hoses, at one moment. A far tap opens up slowly, a leaky hose keeps easing back, Wi-Fi and busy neighbours narrow the flow, and someone downloading at home takes their share. Change any of those and the number changes. Your line is the same; the race is different. And if the number on your download is about one eighth of your plan, that's not a race at all: it's megabytes.
Related lessons: why the internet slows down in the evening · why the game lags when someone starts a download · what really kills your Wi-Fi in the bedroom · cable under the sea or satellite in space · why copying to a USB stick is slow.
Grown-up help on Tooltern: why speed tests give different results on each site, Mbps vs MB/s explained, what counts as good ping, jitter and packet loss, internet slow in the evening, which Ethernet cable you need, the broadband speed calculator and the unit converter for megabits and megabytes. Glossary: slow start, packet loss, latency, throughput.