Also known as: Stored-program architecture

The von Neumann architecture is the stored-program computer design in which both instructions and data live in the same memory, accessed by the processor over a common bus.¹

Overview

In this model a computer has a CPU (with a control unit and an arithmetic/logic unit), a single main memory holding both program and data, and I/O — and the processor works by repeatedly fetching an instruction from memory, decoding it, and executing it. Storing the program in the same memory as the data, rather than wiring it as fixed hardware, is what makes a computer general-purpose: change the program, change what the machine does.

Where it fits

Almost every mainstream computer follows this design, which is why concepts like the instruction set architecture and the fetch-execute cycle are universal. Its famous weakness is the von Neumann bottleneck: the shared path between CPU and memory limits throughput, which is exactly why cache memory and wide modern buses exist. The streaming decode loop in GopherTrunk is, at heart, a von Neumann machine pulling samples and instructions from memory in lockstep.

Sources

1. Von Neumann architecture — Wikipedia, on the stored-program model and the von Neumann bottleneck. ↩

See also

• Central processing unit (CPU)
• Random-access memory (RAM)
• System bus
• Instruction set architecture (ISA)
• Cache memory
• Logic gate