A non-volatile memory that holds magnetic charges in tiny ferrite cores. The direction of the flux determines the 0 or 1. Developed in the late 1940s by Jay W. Forrester and Dr. An Wang, it was used extensively in the 1950s and 1960s as the computer's internal workspace until it was replaced with semiconductor memory.
The cores (bits) are written by sending pulses with half the maximum current down each of the X and Y wires that intersect at that core location. The direction of the pulses determines the magnetization (0 or 1). The core that receives both pulses is changed because it gets more than half the current. Reading is accomplished by a "destructive" read. A "0" is written to the core, and if the core was originally a 0, the sense wire receives no current. If it was originally a 1, current is sensed. The final step to complete the read is to rewrite the bit.
In 1956, IBM paid Dr. Wang USD $500,000 for his patent on core memories, which he used to expand his company, Wang Laboratories. See MRAM
and early memory
Core storage is why the internal workspace of the computer is called "memory." The cores held their content without power. It did remember.
Cores from the Whirlwind Computer
In 1952, this Whirlwind I core plane held 256 bits. Today's memory chips hold billions of bits in that same amount of space. (Images courtesy of The MITRE Corporation Archives.)