1.4. Prefixes for Binary Multiples

For certain topics, especially architecture, computer science is a quantitative science. We shall introduce appropriate discussions of number systems and related matters as they are needed. The first such discussion here treats the matter of scale factors and abbreviations for “big” numbers.

Treating 2¹⁰ (1024) and 10³ (1000) as synonymous introduced a sloppy situation for nomenclature in the fields of computer science, data processing, and data transmission. People were using the prefix kilo, from the International System of Units (SI), not only for its proper meaning of precisely 1000, but also to denote the binary multiple 1024. Doing so introduces an uncertainty of 2.4% in magnitude. The uncertainty compounds to more than 4.8% for the SI prefix mega, and the discrepancy just gets worse and worse for larger quantities. Moreover, kilo is typically abbreviated as a capital K in computer-related contexts, even though the proper SI abbreviation is a small k for kilo (which is an exception to the capitalization of most other abbreviated SI prefixes that denote multiples).

In 1998, the International Electrotechnical Commission (IEC) approved a new standard of names and symbols for the prefixes for binary multiples. The new names have mnemonic analogies to corresponding decimal multiples, as shown in Table 1-2.

The IEC-approved prefixes are not part of the International System of Units (SI), but they have been supported by the International Committee for Weights and Measures (CIPM) and the Institute of Electrical and Electronics Engineers (IEEE^®).

We use these new prefixes in this book. For instance, in the next three tables of this chapter, you will see KiB and MiB denoting kibibytes and mebibytes, where you might have expected KB and MB denoting kilobytes and megabytes, respectively.