📦 Storage Unit Converter — Bytes, KB and KiB Side by Side, Plus the Real Size of a Disk

One number, two answers

A hard drive sold as 1 TB is 10¹² bytes: manufacturers use the decimal prefixes, where kilo-, mega- and giga- mean 1000, 10⁶ and 10⁹ exactly as they do in physics. The byte count does not change, but a program that divides by 2³⁰ and still prints “GB” reports 931. Nothing went missing - that figure is 931 GiB, the binary unit, and the two only look contradictory because different divisions share one label.

Both readings are printed side by side above, since that is the whole answer: “how big is this file” is really “how many bytes”, and the unit is a choice you make afterwards. The gap grows with size:

In practice: drives and SSDs are sold decimal, RAM is binary, Linux ls -lh prints binary, Windows shows a byte count wearing a decimal label, and network speeds are always decimal bits - so a “10 Mbps” line moves at most 1.25 MB/s. That factor of 8 explains most “my download is eight times slower than advertised” reports before any overhead is counted.

How to write the input

A number, an optional space, then a unit. The parser is deliberately narrow:

• Bare number → bytes. 2048 is 2048 B.
• A single b → one bit; a single B → one byte. 1000b is 125 B.
• Sizes ignore case: mb, MB and Mb all mean megabytes, and kb, kB, KB all mean 1000 bytes. Strict SI would read Mb as megabits, but people type it both ways, so a size is read the way file sizes are meant.
• An i before the B selects the binary unit: MiB is 1048576 bytes (Mib, MIB too). No i means decimal.
• Spelled-out bit forms always mean bits: bit, Mbit, Gbit.
• Rates are the exception - there the letter case matters, because it is the only thing distinguishing them: MB/s is bytes, Mbps is bits, Gib/s is binary gigabytes per second. A rate written with no marker at all (100M/s) is read as bits, which is how links are quoted.
• A per-second unit in the size box is refused, not silently shortened: 100 Mbps there would mean “100 megabits”, a different kind of quantity from the one being asked for. The error points at section B.
• Scientific notation and digit grouping are fine: 8e3 KB, 1,024 MB, -3.5 GiB.

Where the transfer time comes from

bytes × 8 ÷ bits-per-second is the entire calculation, which is exactly why it is labelled ideal. Real transfers run below the stated rate for reasons no converter can know about your link: TCP spends the first round trips ramping its congestion window, TLS handshakes and redirects cost a round trip each, Wi-Fi negotiates a rate it cannot sustain, a “shared” 1 Gbps plan is genuinely shared, a disk can cap out well below the bus speed, and mobile networks throttle bulk transfers while a latency test still looks fine.

The second cell applies a flat 70%, which is roughly where a saturated link on a stable connection lands, just to turn “1 min 49 s” into something nearer “2 min 33 s” and make clear which of the two was optimistic. Treat both as lower bounds. If the actual transfer takes more than about twice this, the link is worth investigating rather than the arithmetic.

The comparison line

“About 3.4 × DVD-5” uses nominal media capacities: a 1.44 MB floppy is 1474560 bytes (1440 KiB, itself a mix of conventions), a data CD holds 700 MiB, a DVD-5 4.7 × 10⁹ bytes, a single-layer Blu-ray 25 × 10⁹. Blank media reserve room for lead-in and error correction, so what you can actually write is a few percent under the label - these are the marketing numbers, meant as a feel for size rather than a burn plan.

Questions

Q: Should I write MB or MiB? For a file’s size in a bug report, use whatever your system shows and say which system showed it. For capacity you are buying, decimal. For link speeds, decimal. MiB is only worth typing when the difference changes a decision - a partition table, an LVM extent, a chunk size.

Q: Why does a bare number mean bytes rather than kilobytes? Byte is the only unit with no division hidden in it, so it is the one reading that cannot be wrong. Every other choice loses information; the exact count is in the first cell.

Q: How exact are these figures? Byte counts are integers up to 2⁵³ (about 9 PB), which a double holds exactly, and above that the tool says so instead of pretending. Unit rows are printed with a fixed number of digits, so 1/3 GiB shows as 0.333 GB rather than a full float.

Q: Does what I type get sent anywhere? Parsing, dividing and formatting are arithmetic inside the script: no eval, and no server is asked to do the maths. What you enter does not leave this device.