Mbps vs MB/s: why your download looks eight times slower

Why the same connection shows two numbers

Nothing is wrong when your 300 Mbps plan shows a download at 35 MB/s. The two figures measure the same flow of data in different units:

• Providers and speed tests count bits. A connection sends a stream of bits down a cable or through the air, so networking has always measured speed in bits per second. Plans, broadband labels and speed test results all use Mbps or Gbps.
• Apps count bytes. Files are stored and sized in bytes, so software that writes a file to disk reports how many bytes per second arrive: kB/s, MB/s or GB/s.

The capital letter is the whole difference. b is a bit, B is a byte, and 1 B = 8 b. So 1 MB/s = 8 Mbps, and 300 Mbps ÷ 8 = 37.5 MB/s, the most that plan can show in a download app before any overhead. Each unit has its own primer: what is Mbps and what is MB/s.

Reading the plan table

The table above puts each plan tier next to the most it can show in a download app, a typical figure at about 90%, and how long 1 GB and 50 GB take at that typical rate. Three things are worth reading out of it:

• The typical column is your benchmark. A wired download that reaches it is healthy. On a 300 Mbps plan that is about 33.8 MB/s; seeing 34 MB/s in a launcher means the plan is delivering, not that you are getting a ninth of it.
• The 50 GB column is where plan tiers feel different. At the typical rate, 50 GB takes about 1 h 14 min on 100 Mbps, 24 min 41 s on 300 Mbps and 7 min 24 s on 1 Gbps. For a 1 GB file the same plans differ by well under two minutes.
• The 2 Gbps row assumes a faster-than-gigabit port. Through an ordinary Gigabit Ethernet port, any plan above 1 Gbps looks like about 117.5 MB/s to a single device.

For values between the rows, use the Mbps to MB/s converter, which works in both directions.

Where the missing 6–10% goes

Even a perfect connection never shows the full plan speed ÷ 8 in a download window, for reasons that are easy to quantify:

• Packet headers. On Ethernet with standard 1,500-byte packets and TCP timestamps, each packet carries 1,448 bytes of file data out of 1,538 bytes on the wire: 94.1%. That alone turns 1,000 Mbps into about 941 Mbps, or 117.7 MB/s.
• Wi-Fi. Radio links spend extra time on acknowledgements and retries, and share airtime with every other device, so the share lost is larger and varies from minute to minute.
• Short downloads. A transfer starts slowly and ramps up, so a file that finishes in a second or two never reaches full speed. Judge speed on a download that runs for at least 20–30 seconds.
• MiB/s labels. A tool that counts 1,048,576-byte megabytes shows about 4.6% less for the same speed; this is labelling, not loss.

That is why the typical column uses 90%: headers take about 6%, and a few percent more for real-world conditions is normal on a good line.

Reading the numbers on common screens

• Game launchers and browsers usually show MB/s. Multiply by 8 to compare with your plan: a launcher at 55 MB/s is moving 440 Mbps. Steam has its own quirks, covered in Steam download speed.
• Speed test sites show Mbps, the same unit as your plan, so the result can be compared directly. The speed test checker does that comparison for you.
• Router and system monitors vary. Many show kbps or Mbps for network traffic, while file-copy windows on the same computer show MB/s. Check the letter case before comparing.
• Some tools use binary units such as MiB/s (mebibytes per second), sometimes still labelled "MB/s". 1 MiB/s = 8.388608 Mbps, about 5% more than 1 MB/s, so a reading from such a tool converts to slightly more Mbps.

Labels such as "Mb/s" and "Mbit/s" mean the same as Mbps. "MBps" means megabytes per second, but it is easy to misread, which is why this site writes MB/s.

Is my download slow? A quick check

Take the MB/s figure from your download, multiply by 8, and divide by your plan speed:

share of plan = (MB/s × 8) ÷ plan Mbps
55 MB/s on a 500 Mbps plan = 440 ÷ 500 = 88%

88% on a wired computer is a healthy result. If the same sum gives 30% or 40%, the bottleneck may be Wi-Fi, a busy line or the server you are downloading from. A single download is only as fast as the server sending it, so one slow download proves little; a speed test to a nearby server is the better check.

The reverse sum is useful too. To know what reading a plan should give, divide by 8: on a 1 Gbps plan, a wired test typically tops out around 940 Mbps because of Ethernet overhead, which is 117.5 MB/s in a download app. For MB/s readings of any size, the MB/s to Mbps converter does the multiplication.

The costly mix-up: sizing a plan in the wrong unit

The factor of 8 matters most when you buy something. If a workflow needs files to arrive at 50 MB/s, for example pulling large video projects from a cloud drive, that is 400 Mbps, and with about 10% overhead a 500 Mbps plan is the realistic minimum. Reading 50 MB/s as 50 Mbps would lead you to a plan with an eighth of the capacity. The reverse mistake is more common and more expensive: seeing a download at 35 MB/s on a new 300 Mbps plan, assuming it is broken, and upgrading, when 35 × 8 = 280 Mbps is 93% of the plan.

Bits per second is the standard unit for data transmission in networking equipment, standards and regulation, which is why providers use it (the FCC's 2024 broadband benchmark is 100 Mbps down and 20 Mbps up). The bigger number is a side effect, not a trick. If the idea of bits and bytes is itself the confusing part, bits vs bytes explains where the 8 comes from; for rates in any unit, from B/s to Gbps, the bytes to Mbps converter shows them all at once.