KB, MB, GiB: decimal and binary data units
There are two complete sets of data units and both are correct in their own domain. A kilobyte is 1,000 bytes, a kibibyte is 1,024 bytes, and a 1 TB drive holds 1,000,000,000,000 bytes — which a binary-based file manager reports as 931 GiB. Once you know which base a label uses, every apparent discrepancy resolves to a factor of 1.024, applied once per prefix step.
Where the two systems came from
SI prefixes have meant powers of ten since 1795: kilo is 1,000, mega 1,000,000, giga 1,000,000,000. Computer memory, on the other hand, is addressed in powers of two, so early programmers used 1,024 — the nearest power of two to 1,000 — and borrowed the same prefix names for it. For decades 'kilobyte' meant whichever of the two the author had in mind.
In 1998 the International Electrotechnical Commission created unambiguous binary prefixes: kibi (Ki), mebi (Mi), gibi (Gi), tebi (Ti), each a power of 1,024. IEC 80000-13, published in 2008, states the rule plainly: kB is 1,000 bytes, KiB is 1,024 bytes. The standard is not controversial; the old habit is simply hard to remove from interfaces that were written before it existed.
The arithmetic, prefix by prefix
Each step up multiplies by 1,000 in the decimal set and by 1,024 in the binary set. The two drift apart quickly: after one step the difference is 2.4 %, after three steps (giga) it is 7.4 %, and after four steps (tera) it is 10 %.
| Unit | Bytes | Nearest binary unit |
|---|---|---|
| kB (kilobyte) | 1,000 | 0.9766 KiB |
| KiB (kibibyte) | 1,024 | 1 KiB exactly |
| MB (megabyte) | 1,000,000 | 0.9537 MiB |
| MiB (mebibyte) | 1,048,576 | 1 MiB exactly |
| GB (gigabyte) | 1,000,000,000 | 0.9313 GiB |
| GiB (gibibyte) | 1,073,741,824 | 1 GiB exactly |
| TB (terabyte) | 1,000,000,000,000 | 0.9095 TiB |
| TiB (tebibyte) | 1,099,511,627,776 | 1 TiB exactly |
The useful shortcut: to go from a decimal label to the binary number a file manager will show, divide by 1.024 once per prefix step. A 500 GB drive shows 500 ÷ 1.024³ = 465.7 GiB. A 16 GB memory card shows 14.9 GiB, which is why formatted capacity never matches the printed capacity.
What each vendor actually means
Storage manufacturers use decimal units, because SI prefixes mean powers of ten and because the number on the box is then the number of bytes in the device. A 1 TB SSD contains 1,000,000,000,000 bytes. That is not marketing trickery; it is the definition the industry standardised on.
Memory modules are the opposite. A 16 GiB DIMM holds 17,179,869,184 bytes, and the module is built from chips that are addressed in powers of two. Labelling it 17.2 GB would be technically accurate and practically useless, since nothing in the address space would line up.
Cloud storage and bandwidth quotas almost always use decimal units, because they are billed as a quantity of bytes transferred or stored. Operating systems are the inconsistent ones: macOS switched its file size display to decimal in 2009, Windows still divides by 1,024 while printing 'KB' and 'GB'. Two screenshots of the same file can therefore show two different numbers, and neither is broken.
Network speeds are bits, not bytes
Every network specification is in bits per second, because that is what the physical layer carries. A 100 Mbit/s connection delivers 100,000,000 bits per second, which is 12.5 MB/s — divide by 8, and keep dividing by 1,000 rather than 1,024 if you want the decimal figure.
Real throughput is lower than the nominal rate. A gigabit Ethernet link typically moves about 940 Mbit/s of payload once framing overhead is removed, which is roughly 118 MB/s. A 100 Mbit/s link is usually closer to 90 Mbit/s in practice. When a download manager reports 11 MB/s on a '100 megabit' line, nothing is wrong: that is 88 Mbit/s, which is what the link actually delivers.
A checklist when capacity matters
- Decide whether the number you were given is decimal (kB, MB, GB, TB) or binary (KiB, MiB, GiB, TiB).
- Convert to bytes once, then into whatever unit you need. Two conversions in a row invite a wrong exponent.
- Expect formatted capacity to be lower than the printed figure even after the base conversion, because file systems reserve metadata space.
- For backup planning, use decimal terabytes and assume 90 % usable after formatting and the filesystem.
- Never compare a network speed in Mbit/s with a file size in MB without dividing by 8 first.
Frequently asked questions
Is a kilobyte 1,000 or 1,024 bytes?
1,000 bytes, according to the SI prefixes and IEC 80000-13. The value 1,024 bytes is a kibibyte, written KiB. Both conventions are still in use, which is why the distinction is worth stating explicitly in any document that depends on it.
Why does my 1 TB drive only show 931 GB?
It does hold 1 TB. The file manager divides 1,000,000,000,000 bytes by 1,024 three times and gets 931.32, then labels the result GB. The hardware is exactly as advertised; only the label uses the wrong prefix.
How long should a 1 GB download take on 100 Mbit/s?
1 GB is 8,000,000,000 bits, so 80 seconds at the nominal rate. Real connections deliver 60 to 90 % of nominal throughput, so plan on 90 to 130 seconds.
The data converter lists all twelve units with both symbols, so a decimal label and a binary value can be compared directly.
Related guides
- Unit conversion for cross-border shopping and shipping
- Metric vs imperial: the mistakes that cost real money
Written by the UnitBridge team. Last updated November 19, 2026.