| Bits (bit) | Pebibits (Pibit) |
|---|---|
| 1 Bit | 8.881784197 × 10-16 Pibit |
| 2 Bits | 1.7763568394 × 10-15 Pibit |
| 3 Bits | 2.6645352591 × 10-15 Pibit |
| 4 Bits | 3.5527136788 × 10-15 Pibit |
| 5 Bits | 4.4408920985 × 10-15 Pibit |
| 10 Bits | 8.881784197 × 10-15 Pibit |
| 20 Bits | 1.7763568394 × 10-14 Pibit |
| 25 Bits | 2.22044604925 × 10-14 Pibit |
| 50 Bits | 4.4408920985 × 10-14 Pibit |
| 100 Bits | 8.881784197 × 10-14 Pibit |
| Reference | Bits (bit) | Pebibits (Pibit) |
|---|---|---|
| A plain text message (160 characters) | 1280 bit | 1.13687 × 10-12 Pibit |
| A three-minute MP3 | 24000000 bit | 0.0000000213163 Pibit |
| A smartphone photo | 32000000 bit | 0.0000000284217 Pibit |
| A high-definition film | 3.2 × 1010 bit | 0.0000284217 Pibit |
| A dual-layer Blu-ray disc | 4 × 1011 bit | 0.000355271 Pibit |
The bit is the fundamental unit of information. Its symbol is bit, and its name is a contraction of binary digit, coined by the statistician John Tukey and put into print by Claude Shannon in his 1948 paper A Mathematical Theory of Communication, the work that founded information theory.
A bit is the amount of information carried by a single choice between two equally likely possibilities. A coin landing heads or tails, a switch open or closed, a voltage high or low: each of those settles one bit. That definition is what makes the bit a unit rather than a mere convention of notation. It measures how much uncertainty an answer removes, and it does so in a way that is independent of what the question was about.
Shannon's insight was that this could be counted. A message drawn from an alphabet of thirty-two equally likely symbols carries five bits per symbol, because thirty-two is two to the fifth. If the symbols are not equally likely — as letters in English are not — the average drops, and that gap between the naive count and the true average is exactly what compression exploits. A well-compressed file is one from which the redundant bits have been removed.
In hardware the bit is a physical state: a charge trapped on a floating gate in flash memory, the direction of magnetisation of a domain on a hard disc platter, a pit or land on an optical disc, a pulse of light present or absent in a fibre. All of these encode the same abstract quantity, which is why data can move between them without loss.
Bits are almost never counted singly in storage. They are grouped into octets of eight, and storage capacity is quoted in octets or their multiples. Transmission is different: network and interface speeds are quoted in bits per second, so a connection described as 100 megabits per second delivers about 12.5 megaoctets per second. Confusing the two is the commonest arithmetic error in the whole field.
Where single bits do get counted is in specifications of precision and range. A colour channel with 8 bits holds 256 levels; one with 10 bits holds 1,024. Audio at 16 bits per sample has about 96 decibels of dynamic range, and at 24 bits about 144. A 64-bit address can name about 18 quintillion locations. In every case, each added bit doubles what can be distinguished.
One bit equals 0.125 octets, 0.001 kilobits, or about 0.0009766 kibibits.
The pebibit is a unit of digital information equal to 1,125,899,906,842,624 bits, which is 1,024 tebibits or two to the fiftieth power. Its symbol is Pibit. It is the binary counterpart of the petabit, and the gap between them has widened to 12.6 per cent.
That widening is worth watching, because it is the whole argument for the IEC prefixes in one number. Each step up the ladder multiplies the discrepancy by 1.024, so what began as a harmless 2.4 per cent at the kibibit is an eighth of the quantity here and a fifth at the yobibit. A convention that was reasonable for small numbers becomes indefensible for large ones.
A pebibit is 140,737,488,355,328 octets, or 128 tebioctets. Storage of this size exists in supercomputer memory systems, in the fastest scientific data buffers and in the aggregate memory of very large clusters. The largest machines on the list of the world's fastest computers have main memory measured in pebioctets, and the ability to hold an entire simulation in memory rather than on disc is what makes certain calculations feasible at all.
Because these systems are addressed in binary, the binary unit is the correct one and the decimal figure would be an approximation. A cluster with a pebibit of memory does not have a petabit; it has 12.6 per cent more, and a scheduler that allocated on the decimal figure would leave that much unused.
The unit also appears in filesystem limits. Several widely used filesystems have maximum volume or file sizes expressed as exact powers of two, and where those limits fall in this range they are naturally written in pebibytes or pebioctets. Documentation that converts them to decimal units loses the property that made them memorable.
In everyday computing the pebibit is never encountered. It belongs to specifications, to system architecture and to the design of the largest machines, and a converter needs it for exactly those documents. Anywhere the underlying quantity is a power of two, this is the unit that says so without rounding.
One pebibit equals 1,024 tebibits, 140,737,488,355,328 octets, or about 1.126 petabits.