Convert kB to B

A kilobyte is 1,000 bytes when a standards body defines it and 1,024 when an operating system does. This page uses 1,000; the kibibyte is the unit that means 1,024.

A byte is eight bits, though that was not always fixed — early machines used six, seven or nine. The eight-bit byte won because it holds one character of text and divides neatly in half.

Going from kilobytes to bytes moves the decimal point 3 places to the right and changes nothing else. There is no factor to remember and no rounding to decide: the digits stay in the same order and only their position changes.

1,234 kB is 1234000 B — same digits, moved along. That is worth knowing because it is the one kind of conversion you can check at a glance: if the digits of the answer are not the digits you started with, something other than the conversion has happened to them.

Decimal makes this the simplest conversion in the data units: kilo is 10³, so a kilobyte is 1,000 bytes and 47 kB is 47,000. That is the SI meaning of the prefix, the meaning used by every network standard, and the meaning behind the capacities storage is sold at. The digits stay put and only the decimal point moves.

The binary reading gives 1,024 bytes, a unit properly called the kibibyte and written KiB. The two are 2.4 per cent apart, which is the narrowest gap between any decimal and binary pair — it widens to 4.9 per cent at mega, 7.4 at giga and 10 at tera. That narrowness is why the ambiguity has survived here longest: at kilobyte scale the difference is almost never large enough for anyone to notice they were using the wrong one.

The case of a single letter is worth eight times more than the entire 1,000-against-1,024 question. B is a byte, b is a bit, and there are eight bits in a byte, so 1 kB is 8 kb and 100 kB is 800 kb. Files, memory and storage are counted in bytes; line speeds, bitrates and anything measured per second on a wire are counted in bits, by a convention that goes back to serial links and has never changed.

The practical consequence is that a figure with a per-second attached to it needs dividing by eight before it can be compared with a file. A 256 kbit/s audio stream is 32 kB every second, which is 115 MB an hour. A 64 kbit/s link moves 8 kB a second. Reading either as bytes overstates the throughput eightfold and produces plans that cannot work.

The one place most people meet this conversion daily is a web performance budget: a limit of so many kilobytes of JavaScript, CSS or total page weight, checked against a transfer size a tool reports in bytes. Those budgets are decimal, because they are round numbers chosen by people and compared against network measurements that use powers of ten. A 170 kB script budget is 170,000 bytes.

What matters more than the divisor is which size is being measured. The number that counts against a budget is the compressed transfer size, because that is what crosses the network, and for JavaScript that is typically a third of the file on disk. A bundle of 520,000 bytes uncompressed can arrive as 160,000 bytes of brotli and pass a 170 kB budget comfortably — so a comparison between an uncompressed figure and a budget is not a comparison at all.

Below a megabyte, the numbers that shape everything are powers of two rather than powers of ten. A disk sector is 512 bytes or 4,096; a memory page is 4,096 on nearly every current system; a filesystem block is usually 4,096; a TCP segment on ordinary Ethernet carries up to about 1,460 bytes of payload. None of those is a round number of kilobytes in either system, and all of them decide behaviour that unit conversion does not.

That is why a 1 kB file occupies 4,096 bytes and a 5 kB file occupies 8,192. It is also why buffer sizes are chosen as 4,096, 8,192 or 65,536 rather than 4,000 or 50,000: a buffer aligned to a page or a block moves in whole units and one that is not straddles two. At this scale the useful instinct is not "which kilobyte" but "which boundary is this number near".

A limit expressed in kilobytes on text is a limit in bytes, and UTF-8 does not spend one byte per character. Plain ASCII does cost one, but accented Latin letters and Greek or Cyrillic cost two, most CJK characters cost three, and emoji and many symbols cost four. So a 2 kB field holds 2,000 English characters, around 1,900 characters of French with accents kept, and 500 emoji.

The distinction shows up wherever a limit is enforced somewhere other than where it was declared. A database column may be declared in characters while an index, a protocol field or a cookie limit further along the path counts bytes, and a value that fits by one measure is rejected by the other. Where a limit is being checked against a measurement in bytes, the safe assumption is that everything downstream counts bytes too.

A kilobyte is roughly 1,000 characters of plain English, which is a long paragraph or a short email body. A minimal HTTP request with its headers is one to two kilobytes before cookies, and cookies push it further; a small JSON API response is a few kilobytes; a web font subset for one weight is 15 to 30 kB; a favicon is a few kilobytes; a typical CSS file after compression is tens of kilobytes.

Holding that scale in mind is what the conversion is actually for. When a tool reports 47,318 bytes, calling it 47 kB immediately places it — larger than a request, smaller than a font file, a plausible size for a compressed stylesheet. The precise divisor is irrelevant to that judgement, which is the honest reason the kilobyte ambiguity has never been worth arguing about at this end of the scale.

One test settles it. Take a value that is exactly 1,024 bytes: a tool reporting it as 1.02 kB is decimal, and one reporting it as 1.0 KB is counting in kibibytes whatever letters it prints. On Unix, du and df are binary by default and accept a flag for powers of ten; browser network panels and most web performance tooling report decimal, matching the way transfer rates are quoted.

If the two figures you are comparing differ by about 2.4 per cent and no other explanation fits, the divisor is the explanation. If they differ by a factor of four on a small file, it is block allocation. If they differ by roughly three to one, one of them is compressed and the other is not. Those three checks account for nearly every disagreement at kilobyte scale, and they are worth running in that order.