KB vs KiB: Decimal and Binary Sizes Explained

Decimal (SI) units count in powers of 1000. One kilobyte (KB) is 1000 bytes, one megabyte (MB) is 1,000,000 bytes, one gigabyte (GB) is 1,000,000,000 bytes, and it continues up through terabytes and petabytes. Binary (IEC) units count in powers of 1024: one kibibyte (KiB) is 1024 bytes, one mebibyte (MiB) is 1,048,576 bytes, and one gibibyte (GiB) is 1,073,741,824 bytes.

The binary names with an 'i' — kibi, mebi, gibi — were introduced in 1998 by the IEC specifically to remove the ambiguity between the two systems. They are precise and unambiguous, but they never fully caught on in everyday speech, which is why the same three letters (KB, MB, GB) are still used loosely for both meanings.

Storage manufacturers advertise capacity in decimal units, because that is the SI standard and the numbers look bigger. A '1 TB' drive therefore holds exactly 1,000,000,000,000 bytes. Many operating systems, however, historically report free space in binary units while still writing the label 'GB'. Divide 1,000,000,000,000 by 1,073,741,824 and you get about 931 — so the drive shows as roughly 931 GB, which is really 931 GiB.

Nothing is lost; the same bytes are simply being divided by a larger unit. The gap also widens as sizes grow: it is about 2.4% at the kilobyte level, roughly 5% at the megabyte level, and nearly 10% at the terabyte level. That is why the discrepancy is barely noticeable on small files but obvious on big drives.

As a rule of thumb, storage and networking marketing use decimal units, while memory (RAM) uses binary. Windows reports binary sizes but labels them 'KB/MB/GB'; macOS and Linux switched to decimal for disk sizes years ago, so a drive shows closer to its labelled number. RAM is always counted in binary because of how memory addressing works — that is why you buy 8 GB or 16 GB sticks, never 10 GB.

When a number really matters, check whether it means 1000 or 1024, and prefer the explicit 'i' units — KiB, MiB, GiB — whenever you want to be unambiguous. A quick conversion to plain bytes settles any argument, because bytes are the common ground both systems are built on.

The unit mismatch explains most of the gap, but not all of it. Formatting writes filesystem metadata — inodes, allocation tables, journals — before a single file exists. `ext4` additionally reserves five per cent of the volume for the root user by default, so that a full disk does not stop a server from booting. A laptop usually arrives with a recovery partition and a preinstalled operating system on top of that.

Then every file rounds up to a whole allocation cluster — commonly 4 KiB — so a folder of ten thousand one-byte files occupies about 40 MiB rather than 10 KB. Solid-state drives take one more slice you never see: controllers hold back somewhere between seven and twenty-eight per cent of the flash for wear levelling and garbage collection, which is why enterprise drives are often sold as 960 GB rather than the 1 TB their chips add up to.

A 16 GB memory module genuinely is 16 GiB. Addressing works in powers of two, and a chip cannot be built with a round decimal capacity, so every figure in the memory hierarchy is binary by construction: 64-byte cache lines, 4 KiB pages, 2 MiB huge pages. Nobody sells a 16.78 GB stick, which is what an honest decimal label would say.

Networking went the other way and never looked back. A 300 Mbps link is 300 000 000 bits per second, decimal, and bits rather than bytes — a ceiling of about 37.5 MB/s before any protocol overhead. Keeping three facts straight covers almost every confusing number you will meet: storage is decimal on the box and binary in the operating system, memory is binary everywhere, and links are decimal bits.