A reader once asked us a question we have heard a hundred times since. He paid for 1.5 megabit DSL, the speed test confirmed 1.5 megabits, and yet no download ever went faster than about 180. He wanted to know what was broken. The honest answer: nothing. Both numbers were correct, and they were the same speed wearing different clothes.
That answer confuses people for one simple reason, and it is worth thirty seconds to clear it up for good.
Megabits are not megabytes
Speed tests and internet plans measure in megabits per second, written Mbps. Download windows in your browser measure in megabytes per second, written MB/s. A byte is eight bits. So the number your download shows should be roughly your test result divided by eight.
That reader’s 1.5 megabits divided by eight is about 187 kilobytes per second. His “slow” 180 was his connection running at full speed. The same math today looks like this:
- A 25 Mbps plan downloads files at about 3 MB/s.
- A 100 Mbps plan downloads at about 12 MB/s.
- A 300 Mbps plan downloads at about 36 MB/s.
- A gigabit plan downloads at about 110 to 118 MB/s.
If your download speed times eight lands near your test result, your connection is fine and you can stop reading. If the gap is bigger than that, keep going.
The test measures one short trip. A download measures the whole road.
Our speed test measures the pipe between you and our server for a few seconds, using connections tuned to fill that pipe. It answers one question well: what is this line capable of right now?
A download adds a second party with problems of its own. The server on the far end may cap each downloader to protect itself, may be having a busy night, or may sit on the other side of the world behind a dozen network hops. When one particular site crawls while your speed test flies, the bottleneck is usually their end, not yours. Years ago we settled exactly this argument with a reader by moving the same file he called slow onto a server we controlled; the download promptly ran at his full line speed. The file had never been the problem, and neither had his DSL.
This is also why two speed tests can disagree with each other. Another reader watched one test report more than double what a second test claimed on the same afternoon. Different test servers, different distances, different methods. Trust the test you can repeat: run it several times, at different hours, and watch the pattern rather than any single run.
The quiet bandwidth eaters on your side
When downloads lag the test and the math above does not explain it, something local is usually sharing or shaving the line.
Wi-Fi is the number one suspect. A test over ethernet tells you what your connection can do; a test over Wi-Fi tells you what your Wi-Fi can do. Distance, walls, and neighbors all take a cut before your ISP is even involved.
Background traffic is the number two. Cloud backups, game and system updates, sync folders, someone streaming in the next room. One reader logged on “to view one web page” and watched his usage meter climb several megabytes a minute; his machine was busily updating itself the whole time. When a download matters, pause the rest and watch the difference.
The extras in the path take the remainder. A VPN adds encryption overhead and a detour through its own server. An old router can choke well below a modern plan’s speed. A download that writes to a slow or nearly full disk can stall on the disk, not the network.
The last ten percent belongs to overhead
One more pattern from years of forum questions: the connection that tests at 90 percent of the advertised plan and never one bit more. A gamer once chased the missing tail of his DSL plan for weeks with every tweak tool he could find. He was never going to catch it. Moving data across a network costs some of the line in addressing and packaging, and on many connections that runs 5 to 15 percent. Plans are also sold “up to” a number. Testing a bit under the sticker, consistently, is normal and not worth an evening of registry tweaks.
How to compare fairly, and when to call your ISP
Give the comparison a fair fight: plug in with ethernet, quiet the other devices and downloads, run the test a few times at different times of day, then try a large file from a major, well-connected source. Multiply the download number by eight before judging it.
If wired tests repeatedly land far below your plan at quiet hours, that stops being a math problem and starts being a service problem. Now you have something better than a complaint: you have numbers, dates, and times. Call your provider with those in hand. In our experience that conversation goes very differently than “my internet feels slow.”