What Is a Byte? 8 Bits Today, 1 to 6 When It Was Coined

The first bytes on record could be a single bit long. Werner Buchholz, the IBM engineer who coined the word, wrote in June 1956 that his machine would take characters of 1 to 6 bits, and could “handle bytes of only one bit for logical analysis.”

His memo already treats the word as settled, in the phrase “characters, or ‘bytes’ as we have called them.” An 8-bit maximum followed that September, and the fixed 8 came in the 1960s, after a fight inside IBM in which, by Fred Brooks’s account, he and chief architect Gene Amdahl each quit the company once.

What is a byte?

A byte is a group of binary digits, or bits, that a computer stores and moves as one unit, and today it almost always means 8 bits. Eight bits can be set in 256 different patterns, enough to hold a whole number from 0 to 255 or one character of plain English text, such as A, x or $.

A bit is a single 0 or 1, the smallest unit of storage. Claude Shannon’s 1948 paper “A Mathematical Theory of Communication” calls these units “binary digits, or more briefly bits, a word suggested by J. W. Tukey.”

In the international standard, a kilobyte is 1,000 bytes and a megabyte 1,000,000, though most memory makers use megabyte for 1,048,576 bytes, the size the International Electrotechnical Commission named the mebibyte in 1998.

The first bytes were sized for 1950s machines

Buchholz’s company-confidential memo, dated June 11, 1956, described LINK, the part of IBM’s Stretch computer that connected it to the outside world. In a 1977 letter to BYTE magazine, he explained that the byte’s “first use was in the context of the input-output equipment of the 1950s, which handled six bits at a time.”

Stretch’s memory was then planned around a 60-bit word, the number of bits it moved in one memory cycle. In a memo of July 31, 1956, Buchholz noted that “60 is a multiple of 1, 2, 3, 4, 5, and 6,” so every byte size from 1 to 6 fit a word evenly. The same memo also weighed 64: since 64, unlike 60, is a power of 2, a 64-bit word “would permit a completely binary address arithmetic.”

At a meeting on September 17, 1956, attended by Buchholz, Brooks and four others, the word became 64 bits and the largest byte for input and output became 8 bits. Stretch, renamed the IBM 7030 and delivered to Los Alamos in April 1961, let programs choose any byte size from 1 to 8 bits.

The word came from bite, and the y was deliberate

The first published use, Buchholz wrote, came in a June 1959 paper by Gerrit Blaauw, Brooks and himself, “Processing Data in Bits and Pieces.” In a 1995 letter to BYTE, Louis G. Dooley wrote that the word “was coined around 1956 to 1957” on SAGE, an air-defense project at MIT Lincoln Laboratory, but his letter cites no document.

Planning a Computer System, the 1962 book on Stretch that Buchholz edited, explains the spelling on page 40: “The term is coined from bite, but respelled to avoid accidental mutation to bit.”

The same page defines a byte as the bits that encode one character, or that move in parallel to and from input-output units. Parallel bits travel side by side, where a serial link such as USB sends them one at a time.

Eight bits won the biggest fight of the System/360 project

In IBM’s words, System/360, launched on April 7, 1964, replaced all five of its computer product lines with “one strictly compatible family.” Brooks was its project leader and Amdahl its chief architect.

Brooks told the Computer History Museum in 2007 that two of the 12 or 13 teams in an internal design competition, Amdahl’s and Blaauw’s, came in with “essentially the same concept” and one big difference: Amdahl’s used a 6-bit byte, Blaauw’s an 8-bit byte.

Then, Brooks said, “came the biggest internal fight, and that was between the six and eight bit byte.” He and Amdahl “each quit once that week, quit the company,” until Manny Piore, IBM’s senior scientist, got them back together.

Brooks made the decision, Amdahl appealed to Bob Evans, Brooks’s boss, and Evans confirmed it. Brooks said “the reason was it opened the lower case alphabet.” He saw language processing as a market IBM could not enter while it used 6-bit character sets.

Of all his technical accomplishments, Brooks said, “making the 8-bit byte decision is far and away the most important.”

Amdahl’s own oral history, recorded in 2000, does not mention quitting over the byte. He said he “didn’t want to continue with the 8-bit byte” and wanted 24-bit and 48-bit words instead of 32 and 64, for “a more rational floating point system.”

The 1964 IBM Journal paper by Amdahl, Blaauw and Brooks named a different factor first: “Most important of these factors was coding efficiency,” since numeric data in business records was “more than twice as frequent as alphanumeric,” and the 8-bit plan stored a decimal digit in 4 bits rather than 6.

The 8-bit byte is older than System/360

“Many have assumed that byte, meaning 8 bits, originated with the IBM System/360,” Buchholz wrote in 1977. IBM’s history page says the System/360 architecture “pioneered the 8-bit byte still in use on computers today,” though Stretch’s design had an 8-bit maximum by September 1956.

What System/360 changed, Buchholz wrote, was to fix the byte “at the 8 bit maximum,” for economy, and to number memory byte by byte instead of bit by bit. “Since then the term byte has generally meant 8 bits.”

ASCII fits in 7 bits

ASCII, the American standard code for text, was approved on June 17, 1963, as “the 7-bit coded character set,” with room for 128 characters. Its table left most of two columns unassigned, and the 1967 edition filled them with the lowercase letters.

UTF-8, the text encoding specified in RFC 3629, keeps each ASCII character to one byte and uses two to four for any other. An em dash, U+2014, takes three. Images get packed into bytes too: a GIF stores each pixel as an index of up to 8 bits into a color table, which caps each image at 256 colors.

Octet is the word for exactly 8 bits

RFC 791, the 1981 specification of the Internet Protocol, defines an octet as “An eight bit byte.” The C language still defines a byte by its job, as an “addressable unit of data storage large enough to hold any member of the basic character set,” with at least 8 bits.

When	Where	Bits in a byte
June 1956	Buchholz’s Stretch memo	1 to 6
September 1956	Stretch planning meeting	Up to 8
April 1961	IBM 7030 (Stretch), delivered	1 to 8, chosen by the program
April 1964	IBM System/360	8, fixed
September 1981	RFC 791, Internet Protocol	8, called an octet
2024	C standard working draft	At least 8

The first byte could be a single bit, and its y was there to stop the word itself from turning into bit. The fixed 8 took a 1960s fight in which, by Brooks’s account, IBM’s project leader and chief architect each quit the company once, and the word has generally meant 8 bits ever since. Strictly FYI.