How Do Computers Really Work? Inside CPU, Memory, and Architecture

Introduction

People have created countless machines to solve problems, from Mars rovers to nuclear‑submarine navigation computers. In 1945 John von Neumann proposed the first computing model, and virtually every modern computer follows its principles. This article explores how computers work.

Computer Architecture Basics

A computer is a data‑processing machine composed of a processor (CPU) and memory (RAM). RAM stores both instructions and the data they manipulate, while the CPU fetches instructions and data from RAM to perform calculations.

Memory

Memory is divided into many addressable cells, each holding a small amount of data identified by a numeric address. Accessing a cell involves reading or writing a byte (8 bits) via signal lines that carry binary values (high voltage = 1, low voltage = 0).

Signal lines that transfer the same data form a bus . An 8‑line address bus carries addresses to RAM, while an 8‑line data bus carries the actual bytes. The CPU constantly exchanges data with RAM via these buses.

CPU and Registers

The CPU contains internal registers that hold temporary values and can perform simple arithmetic. Typical operations include copying data between memory locations and registers, and adding values stored in registers.

Copy data from memory address 220 to register 3.

Add the value in register 3 to the value in register 1.

The set of all operations a CPU can perform is called its instruction set . Machine code is essentially a sequence of numbers that represent these instructions and are stored in RAM.

Instruction Execution Cycle

The CPU repeatedly performs a four‑step cycle: (1) fetch the instruction at the address held by the program counter (PC), (2) increment PC, (3) execute the instruction, and (4) repeat from step 1. The PC is initialized to a fixed address at power‑on, often pointing to firmware such as the BIOS.

CPU Architecture Variants

Different CPUs use different architectures, which means they understand different instruction sets. For example, x86 is dominant on desktops, while mobile devices use ARM architectures. This explains why a PlayStation game cannot run on a PC and why iPhone apps cannot run on a Mac.

Early CPUs were 4‑bit (e.g., Intel 4004). Later came 8‑bit, 16‑bit, 32‑bit, and today most CPUs are 64‑bit, allowing them to address vastly larger memory spaces (2⁶⁴ bytes ≈ 170 billion GB).

Endianness

Byte order can be big‑endian (most significant byte first) or little‑endian (least significant byte first). Most modern CPUs use little‑endian, but network protocols often assume big‑endian, requiring conversion when exchanging binary data.

Compilers

Programs are written in high‑level languages and then translated by a compiler into the CPU’s instruction set. The compiler maps complex language constructs to simple CPU operations, enabling billions of instructions to be expressed in a few readable lines of code.

if x == 0
    compute_this()
else
    compute_that()

Operating Systems

Compiled programs must interact with an operating system to perform I/O, access hardware, and manage resources. Because OS APIs differ across platforms, the same binary cannot run on both Windows and macOS even if the CPUs share the same architecture.

Cache Hierarchy

To bridge the speed gap between the fast CPU and slower RAM, modern CPUs include multiple cache levels. Level‑1 cache (≈10 KB) is fastest, followed by Level‑2 (≈200 KB) and Level‑3 (≈several MB). Caches exploit temporal and spatial locality to keep frequently accessed data close to the CPU, reducing RAM accesses dramatically.

Memory Hierarchy

Beyond caches, the memory hierarchy includes RAM (primary storage) and secondary storage (hard disks, SSDs). RAM is fast but limited (1‑10 GB typical), while disks provide large capacity but are orders of magnitude slower. When RAM fills, data is swapped to disk, causing “thrashing" and severe performance loss.

External and Tertiary Storage

Further down the hierarchy are network‑attached storage and archival media (tapes, optical disks). Access times increase from milliseconds for local disks to hundreds of milliseconds for network storage, and minutes or more for tape archives.

Trends

Fast storage technologies (SSD, NVMe) are narrowing the gap between primary and secondary storage, while the cost of slow storage continues to drop. Hybrid drives combine SSD speed for hot data with HDD capacity for cold data, mirroring the CPU’s cache strategy.