IT lexicon › Hardware › Triple modular redundancy (TMR)

Triple modular redundancy (TMR)

Hardware På svenska → Updated: 2026-11-03

Do the same thing three times and let the majority decide — fault tolerance by sheer overprovisioning.

John von Neumann worked through the idea in the 1950s: reliable systems built from unreliable parts. Three identical units compute in parallel and a voter picks whatever answer at least two of them agree on. A fault in one unit is therefore masked outright rather than merely detected. The Saturn V's guidance computer in the 1960s used TMR, and it remains standard in space electronics, where radiation flips bits as a matter of routine.

The cost is at least triple the area and power. The voter becomes a new single point of failure, and three copies of the same bug produce three identical wrong answers — redundancy guards against random faults, not design faults.

← Back to the lexicon