Unlike magnetic drums, data stored in magnetic core memory is volatile, which means that it will disappear when the electricity stops flowing. Core memory is capable of random access from one moment to the next, whereas access to different areas of drum memory can be done on successive revolutions only. A significant improvement over drums derives from the lack of moving parts in a core. Magnetic core memory is the precursor to the integrated circuits of DRAM¹ used in all desktop computers of this era.

Core memory is constructed as a three-dimensional matrix of ferrite toroids². Each ferrite bead sits at the node of three wires threaded through the hole in the torus. Two wires are at right-angle to each other, the third is a sense wire.

Electricity switched through the X and Y wires cause the magnetic polarity of the torus to change accordingly with a corresponding flow of electricity in the sense wire. Direction of electrical flow in the sense wire indicates the magnetic polarity that existed in the torus before it was changed by the read operation. Electronic logic circuitry controls the addressing, reading, writing, and maintenance of data in the core. Binary data in the core is encoded as magnetic polarities. Core memories are usually made in bricks of 4096 words.

Two examples of computers that use core memory are:

• Honeywell DDP516³.
• Redifon R2000⁴.

The DDP516 was used to drive the flight simulation of the Blackburn Buccanneer S2B, part of the United Kindom's nuclear strike force. Each morning the computer required the programme to be loaded into core memory from punched paper tape, whereas the R2000 was loaded from 9-track reel-to-reel magnetic tape.