Data units and their equivalents
| Unit | Symbol | In bytes |
|---|---|---|
| Terabyte | TB | 1,000,000 MB |
| Gigabyte | GB | 1,000 MB |
| Megabyte | MB | 1 MBbase unit |
| Kilobyte | kB | 0.001 MB |
| Byte | B | 0.000001 MB |
| Tebibyte | TiB | 1,099,512 MB |
| Gibibyte | GiB | 1,074 MB |
| Mebibyte | MiB | 1.04858 MB |
| Megabit | Mb | 0.125 MB |
| Bit | bit | 0.000000125 MB |
| Petabyte | PB | 1,000,000,000 MB |
| Pebibyte | PiB | 1,125,899,907 MB |
| Kibibyte | kiB | 0.001024 MB |
| Gigabit | Gb | 125 MB |
Frequently asked questions
Why does my 1 TB drive show 931 GB?
The manufacturer counts in base 10 (1 TB = 1,000 billion bytes), the system in base 1,024. 10¹² divided by 2³⁰ gives 931: no space is missing, only the display unit differs. The correct name for what Windows calls GB is actually gibibyte.
MB or Mb for my connection?
Speeds are advertised in megabits per second (Mb/s), eight times smaller than megabytes. An 800 Mb/s fiber line downloads about 100 MB per second.
Has a byte always been eight bits?
The byte, no; the octet, yes. A byte first meant the group of bits needed to code one character, and its size followed the machine: six bits on CDC mainframes, seven, nine on the 36-bit PDP-10. French coined octet precisely to remove that ambiguity, and network standards use the same word for the same reason when precision matters. IEC 80000-13 settled it: an octet is eight bits, written B in English and o in French.
How much does one character of text weigh?
One to four bytes, depending on the alphabet. In UTF-8, the encoding of nearly the whole web, an unaccented Latin letter takes one byte, an accented letter or a Cyrillic, Greek, Hebrew or Arabic one takes two, a Chinese, Japanese or Korean character and most Indic ones take three, an emoji four. The same thousand characters therefore weigh about 1 kB in English, 1.1 kB in French, 2 kB in Russian and 3 kB in Japanese. Composed emoji go further: a flag or a family joined by zero-width joiners is 8 to 25 bytes on its own.
Why does a 3 kB file take 4 kB on disk?
Because a file system allocates space in blocks, not byte by byte. NTFS, ext4 and APFS use 4 kibibyte blocks by default: a 500-byte file occupies 4,096, and ten thousand such files eat 40 MB to hold 5 MB of real content. The file properties window in Windows shows both figures, size and size on disk. On a memory card formatted in exFAT, where the block often reaches 128 kibibytes, the gap becomes spectacular.
The kilo that means 1024 is not a kilo
An SI prefix always multiplies by a thousand: kilo is 1000, mega a million, giga a billion. Computing borrowed those prefixes for powers of two for decades, so that a kilobyte meant 1024 bytes. The IEC 80000-13 standard, published in 1998, settled it by creating separate binary prefixes: the kibibyte is 1024 bytes, the mebibyte 1024 kibibytes, and so on. The kilobyte went back to being what it should always have been, one thousand bytes.
The gap shows on every memory card bought. A drive sold as one terabyte does hold a thousand billion bytes, but Windows counts them in binary units and displays 931 gigabytes, a third of a terabyte apparently gone. Nothing vanished, two counting systems crossed. The second trap is a single change of case: lowercase b is the bit, uppercase B the byte, and eight bits make one byte. A fibre line advertised at 1 gigabit per second therefore delivers at best 125 megabytes per second.