File Size Converter

Why one word ended up meaning two numbers

Memory is addressed with binary addresses, so the quantities that fall out of a memory chip are powers of two: 1024, 65536, 1048576. When those quantities needed names, the nearest decimal prefix was borrowed — 1024 was close enough to a thousand to be called a kilobyte, and the 2.4% error did not matter. It compounds, though. At mega the two readings differ by 4.86%, at giga by 7.37%, at tera by 9.95%, and by peta you are arguing about an eighth of the total. The gap is (1024/1000)n - 1, and every rung you climb it gets worse.

In 1999 the IEC ended the ambiguity by naming the binary quantities properly: kibi, mebi, gibi, tebi, with the symbols KiB, MiB, GiB, TiB. The standard is settled. The usage is not, because decades of software, documentation and habit predate it. So the practical rule is the one this page is built around: a prefix is a guess, a byte count is a fact. When a number has to be right — a quota, a contract, an upload limit, a bug report — quote the bytes.

Reading the three boxes

Type 1 GB and the page tells you it is 1,000,000,000 bytes, which is 953.67 MiB. Type 1 GiB and it is 1,073,741,824 bytes, which is 1.07 GB. Same box, both directions: a bare number is read as a byte count and expanded onto both ladders; a number with a unit is resolved down to bytes. The unit picker overrides whatever you typed, so you can take one figure and ask “what if that was actually MiB?” without retyping it.

The meter under the results is the honest headline: it shows how far apart the two conventions are at the step your number lands on. Under 1000 bytes it reads zero, because below a kilobyte there is nothing to disagree about.

The 1 TB drive question, done as arithmetic

Type 1 TB above and read the GiB row: 10¹² bytes is 931.32 GiB, or 0.909 TiB. That is the whole of the famous “missing” capacity, and nothing is missing — a program that labels 2³⁰ bytes “GB” is dividing by a bigger number than the one the capacity was quoted in, so it prints a smaller figure for the same drive. Which convention your own system uses is its choice, and this page does not assume: set the unit picker to GiB, compare with what your machine shows, and you will know. Filesystem overhead and reserved space are a separate subtraction on top, typically much smaller, and they depend on the filesystem.

Open it with your number already in it

The converter reads its input from the address, so a link can carry a finished answer. ?q= takes the size text exactly as you would type it, percent-encoded: ?q=500%20GB opens on 500,000,000,000 bytes, which the IEC box shows as 465.66 GiB — the “why does my 500 GB drive show 465 GB” answer, with no click. &unit= forces how the number is read (any value from the picker: B, kB, KiB, MB … PiB, or auto), so ?q=1.5%20TiB&unit=GB reads the 1.5 as gigabytes — the typed TiB is overridden — and shows 1,500,000,000 bytes, about 1.4 GiB. &precision= sets the decimals (0, 1, 2, 3, 4 or 6). A value the page cannot use is ignored — ?q=garbage opens on the usual 1 GB example rather than an error page. As you edit, the address bar follows along without adding Back-button entries, and Copy link copies it.

Bits are a different mistake

The SI-versus-IEC gap tops out around 10%. The bit-versus-byte mix-up is a factor of eight, and it hides in the same sentence: link and network rates are quoted in bits per second, storage in bytes. A 100 Mbit/s connection moves at most 12.5 MB/s, and the only thing distinguishing the two is the case of the b. This converter handles bytes only — every unit on it is a byte unit — so divide a bitrate by 8 before you bring it here.

Honest limits

Byte counts are JavaScript numbers, which are IEEE-754 doubles, exactly as in the library this page ports. Integers above 2⁵³ - 1 — 9,007,199,254,740,991 bytes, about 9.01 PB — cannot all be represented, so beyond that the byte box shows the nearest representable value and says so instead of pretending to be exact. A size that does not land on a byte boundary, such as 1.44 MiB, gives a fractional byte count (1,509,949.44) and the page prints the fraction rather than rounding it away, because a fractional result tells you the figure you were given was already rounded. And the ladder table stops at peta: exa, zetta and yotta exist in both standards, and typing them works, but a table that lists them is padding.

Frequently asked

Does 1 GB mean 1,000,000,000 bytes or 1,073,741,824? Both numbers are real, and they have different names. Giga is an SI prefix and the SI Brochure defines it as 10⁹, so 1 GB is 1,000,000,000 bytes. 1,073,741,824 is 2³⁰, and IEC 80000-13 gives that quantity its own name and symbol: 1 gibibyte, 1 GiB. So the standards answer is unambiguous — GB is the decimal one, GiB is the binary one — and the practical answer is that a label reading GB only tells you which was meant if whoever wrote it was being careful. When it matters, ask for the byte count instead of the prefix; that is the number this page puts in the third box.

What is a gibibyte, and who decided KiB and MiB were words? The IEC published the binary prefixes in 1999, as Amendment 2 to IEC 60027-2; they now live in IEC 80000-13:2008, the standard for quantities and units in information science and technology. The scheme takes the first two letters of the SI prefix and adds “bi” for binary: kibi (Ki) = 2¹⁰ = 1024, mebi (Mi) = 2²⁰, gibi (Gi) = 2³⁰, tebi (Ti) = 2⁴⁰, pebi (Pi) = 2⁵⁰. Attached to a byte they give KiB, MiB, GiB, TiB and PiB. They exist precisely so that a binary quantity does not have to borrow a decimal prefix and hope the reader guesses right.

Why does a 1 TB drive not show 1 TB of space? Start with the arithmetic, which you can reproduce above by typing 1 TB and reading the GiB row: 10¹² bytes divided by 2³⁰ is 931.32, and divided by 2⁴⁰ is 0.909. So a capacity of exactly one decimal terabyte, displayed by a program that labels 2⁴⁰ bytes “TB”, reads as 0.91 — and as 931.32 if that program’s unit is 2³⁰. No bytes went missing; two different units are both being called TB and GB. Which convention any particular operating system or file manager uses is that program’s own choice, and this page does not guess at yours: set the unit picker to GiB, compare with what your machine shows, and you will know which one it means. Formatting overhead and reserved space are a separate, smaller subtraction on top of that, and they vary by filesystem.

Is it kB or KB, and why is the k lowercase? In SI the symbol for kilo is a lowercase k — capital K is the symbol for the kelvin — so 1000 bytes is correctly written 1 kB. The IEC binary prefix does use a capital: kibi is Ki, so 1024 bytes is 1 KiB. That leaves plain “KB” defined by neither standard, which is why you should read it as “about a thousand bytes, one or the other” rather than as a precise quantity. This page follows both standards: it writes kB for 1000 bytes and KiB for 1024, and it accepts whatever you type, in any case, on the way in.

How many bytes are in a MB, a MiB, a GB and a GiB? 1 kB = 1,000 bytes and 1 KiB = 1,024 — a 2.4% difference. 1 MB = 1,000,000 and 1 MiB = 1,048,576 — 4.86%. 1 GB = 1,000,000,000 and 1 GiB = 1,073,741,824 — 7.37%. 1 TB = 1,000,000,000,000 and 1 TiB = 1,099,511,627,776 — 9.95%. The gap is (1024/1000)ⁿ - 1 at step n, so it compounds: every rung of the ladder you climb, the two answers drift a little further apart. That is why a mix-up costs nothing on a text file and roughly a tenth of the number on a multi-terabyte array.

What about megabits — is 100 Mbps the same as 100 MB/s? No, and it is a factor of eight, which dwarfs the 7% this page is about. Network and link rates are quoted in bits per second, and a byte is eight bits, so 100 Mbit/s is 12.5 MB/s at best. The two are told apart by the case of the b: Mb is megabit, MB is megabyte. This converter deliberately handles bytes only — every unit on it, B through PiB and PB, is a byte unit — so if you have a bitrate, divide by 8 before you bring it here.

Does anything I type get uploaded? No. The conversion is a few dozen lines of arithmetic running in this tab, in a file that contains no fetch, no XMLHttpRequest and no upload of any kind, so there is no code path that could send a number anywhere. Nothing is stored either — reload the page and it is back to its example. The page does load one cookieless pageview counter from stats.dankdev.com, which records this page’s address and title and nothing you type into it.

Why does the byte count sometimes have a decimal point? Because you asked for a size that does not land on a byte boundary. 1.44 MiB is 1,509,949.44 bytes, and the page says so rather than rounding quietly — a real file is always a whole number of bytes, so a fractional result is a sign that the size you were given was itself rounded. The same honesty applies at the top end: byte counts here are IEEE-754 doubles, exactly as in the library this page is built on, so above 2⁵³ - 1 bytes (about 9.01 PB) not every integer can be represented and the page labels the figure as the nearest value rather than an exact count.