Here are the SI magnifying prefixes, along with the corresponding binary interpretations in common use:
prefix decimal binary
kilo- 1000^1 1024^1 = 2^10 = 1,024
mega- 1000^2 1024^2 = 2^20 = 1,048,576
giga- 1000^3 1024^3 = 2^30 = 1,073,741,824
tera- 1000^4 1024^4 = 2^40 = 1,099,511,627,776
peta- 1000^5 1024^5 = 2^50 = 1,125,899,906,842,624
exa- 1000^6 1024^6 = 2^60 = 1,152,921,504,606,846,976
zetta- 1000^7 1024^7 = 2^70 = 1,180,591,620,717,411,303,424
yotta- 1000^8 1024^8 = 2^80 = 1,208,925,819,614,629,174,706,176
Here are the SI fractional prefixes:
prefix decimal jargon usage
milli- 1000^-1 (seldom used in jargon)
micro- 1000^-2 small or human-scale (see micro-)
nano- 1000^-3 even smaller (see nano-)
pico- 1000^-4 even smaller yet (see pico-)
femto- 1000^-5 (not used in jargon---yet)
atto- 1000^-6 (not used in jargon---yet)
zepto- 1000^-7 (not used in jargon---yet)
yocto- 1000^-8 (not used in jargon---yet)
The prefixes zetta-, yotta-, zepto-, and yocto- have been included
in these tables purely for completeness and giggle value; they were
adopted in 1990 by the `19th Conference Generale des Poids et
Mesures'. The binary peta- and exa- loadings, though well
established, are not in jargon use either --- yet. The prefix
milli-, denoting multiplication by 1000^(-1), has always
been rare in jargon (there is, however, a standard joke about the
`millihelen' --- notionally, the amount of beauty required to
launch one ship). See the entries on micro-, pico-, and
nano- for more information on connotative jargon use of these
terms. `Femto' and `atto' (which, interestingly, derive not
from Greek but from Danish) have not yet acquired jargon loadings,
though it is easy to predict what those will be once computing
technology enters the required realms of magnitude (however, see
attoparsec).
There are, of course, some standard unit prefixes for powers of 10. In the following table, the `prefix' column is the international standard suffix for the appropriate power of ten; the `binary' column lists jargon abbreviations and words for the corresponding power of 2. The B-suffixed forms are commonly used for byte quantities; the words `meg' and `gig' are nouns that may (but do not always) pluralize with `s'.
prefix decimal binary pronunciation
kilo- k K, KB, /kay/
mega- M M, MB, meg /meg/
giga- G G, GB, gig /gig/,/jig/
Confusingly, hackers often use K or M as though they were suffix or
numeric multipliers rather than a prefix; thus "2K dollars", "2M
of disk space". This is also true (though less commonly) of G.
Note that the formal SI metric prefix for 1000 is `k'; some use this strictly, reserving `K' for multiplication by 1024 (KB is thus `kilobytes').
K, M, and G used alone refer to quantities of bytes; thus, 64G is 64 gigabytes and `a K' is a kilobyte (compare mainstream use of `a G' as short for `a grand', that is, $1000). Whether one pronounces `gig' with hard or soft `g' depends on what one thinks the proper pronunciation of `giga-' is.
Confusing 1000 and 1024 (or other powers of 2 and 10 close in magnitude) --- for example, describing a memory in units of 500K or 524K instead of 512K --- is a sure sign of the marketroid. One example of this: it is common to refer to the capacity of 3.5" microfloppies as `1.44 MB' In fact, this is a completely bogus number. The correct size is 1440 KB, that is, 1440 * 1024 = 1474560 bytes. So the `mega' in `1.44 MB' is compounded of two `kilos', one of which is 1024 and the other of which is 1000. The correct number of megabytes would of course be 1440 / 1024 = 1.40625. Alas, this fine point is probably lost on the world forever.
[1993 update: hacker Morgan Burke has proposed, to general approval on USENET, the following additional prefixes:
