In the days of 8-bit computers, two of the more common memory configurations were 16K and 64K, implemented as eight RAM chips of 16kbit or 64kbit respectively. The setup was one chip per bit over the data bus, so you couldn't use a pair of 64kbit chips to supply 16K of RAM. (Well, presumably you could if you were willing to let memory access take four times as long, but nobody was.)
For example, when the Commodore 64 was released in 1982, it shipped with 64K of RAM, which was at that time unusual and moderately expensive relative to the previously more common 16K, but became the typical configuration as time went on.
About what year did it get to the point where 64kbit RAM chips didn't cost significantly more than 16kbit so that even if you were aiming at the lowest end of the market, it wouldn't make sense to provide less than 64K?