Source : Free On-Line Dictionary of Computing

ferrite core memory
     
         (Or "core") An early form of {non-volatile storage}
        built (by hand) from tiny rings of magnetisable material
        threaded onto very fine wire to form large (e.g. 13"x13" or
        more) rectangluar arrays.  Each core stored one {bit} of data.
        These were sandwiched between {printed circuit boards}(?).
        Sets of wires ran horizontally and vertically and where a
        vertical and horizontal wire crossed, a core had both wires
        threaded through it.
     
        A single core could be selected and magnetised by passing
        sufficient current through its horizontal and vertical wires.
        A core would retain its magnetisation until it was
        re-magnetised.  The two possible polarities of magnetisation
        were used to represent the binary values zero and one.
     
        A third "sense" wire, passed through the core and, if the
        magnetisation of the core was changed, a small pulse would be
        induced in the sense wire which could be detected and used to
        deduce the core's original state.
     
        Some core memory was immersed in a bath of heated oil to
        improve its performance.
     
        Core memory was rendered obsolete by {semiconductor} memory.
     
        For example, the 1970s-era {NCR 499} had two boards, each with
        16 {kilobytes} of core memory.
     
        (1996-03-04)