What Is an Embedded System?

In this lesson:

• what makes a computer "embedded", dedicated function, real-time, fixed resources
• MCU (microcontroller) vs MPU (microprocessor): the core hardware split
• why an MCU boots in milliseconds and an MPU needs external RAM and an OS
• how to choose between them
• the "MPU" naming trap (microprocessor vs memory protection unit)

What "embedded" means

A laptop is a general-purpose computer: it runs arbitrary programs you install. An embedded system is a computer built into a larger product to do one dedicated job, the controller in a washing machine, a motor drive, a pacemaker, a car's ECU. You don't install apps on it; its program is fixed firmware.

Embedded systems typically share these traits:

• Dedicated function, one purpose, decided at design time.
• Real-time constraints, must respond within deadlines (fire the spark plug now, not 5 ms late).
• Fixed, limited resources, kilobytes of RAM, not gigabytes; tight power and cost budgets.
• High reliability, runs unattended for years; often can't crash or be rebooted by a user.
• Direct hardware interaction, talks to sensors, motors, and radios through peripherals, not through a desktop OS.

MCU vs MPU: the fundamental hardware split

The first architectural decision is what kind of processor sits at the center.

A microcontroller (MCU) integrates the CPU core, memory (flash + SRAM), and peripherals onto a single chip. It's a complete computer in one package, add power and a crystal and it runs. Think STM32, AVR, PIC, ESP32.

A microprocessor (MPU), here meaning an application processor, is essentially just the CPU (often with a GPU). It has no usable on-chip program memory or RAM; it needs external DRAM, external flash/eMMC, and a power-management chip, and it runs a full OS like Linux. Think the SoC in a Raspberry Pi, i.MX, or a phone.

        MCU (one chip)                       MPU (a system of chips)
 ┌───────────────────────────┐        ┌─────────┐   ┌──────────┐
 │  CPU core                 │        │  CPU    │──▶│ external │
 │  Flash (program)          │        │ (+ GPU) │   │  DRAM    │
 │  SRAM (data)              │        └────┬────┘   └──────────┘
 │  Peripherals (GPIO,UART,  │             │        ┌──────────┐
 │   timers, ADC, ...)       │             └───────▶│  eMMC /  │
 └───────────────────────────┘                      │  flash   │
   self-contained, boots from               +PMIC, +external storage,
   internal flash in ms                      boots an OS in seconds

The comparison that matters

The rule of thumb: choose an MCU when the job is real-time control, low power, low cost, and instant-on. Choose an MPU when you need an operating system, a rich GUI, networking stacks, a filesystem, or serious compute (video, ML). Many products use both, an MPU for the application and an MCU as a real-time co-processor.

This course is about the MCU world (Cortex-M class), where you own every byte and every cycle.

The line is blurry: "embedded Linux"

The MCU/MPU split isn't absolute. High-end MCUs now have megabytes of RAM and run lightweight Linux; application processors can run bare-metal. "Embedded Linux" sits in the middle, an MPU running Linux but in a fixed-function product (a router, a smart thermostat). The deciding questions are practical: do you need an OS and its ecosystem, and can you afford the external memory, power, and boot time it requires?

The "MPU" naming trap

Watch out: MPU has two meanings in embedded.

• MicroProcessor Unit, an application processor (this lesson).
• Memory Protection Unit, a small hardware block inside many Cortex-M microcontrollers that enforces memory access permissions (covered in memory-protection topics).

So a Cortex-M microcontroller (an MCU) can contain an MPU (memory protection unit). Always read MPU from context; senior engineers disambiguate automatically.

Gotchas

• "MCU vs MPU" is about integration, not just speed. The defining difference is on-chip memory and self-containment, not clock rate. A fast MCU can outrun a slow MPU; the split is about whether memory/OS are integrated.
• MPUs can't boot instantly. They load an OS from external storage into external DRAM, seconds, not milliseconds. If your product must be responsive at power-on (a safety control), that boot time matters.
• Determinism favors the MCU. An MMU, caches, and an OS scheduler make worst-case timing harder to bound on an MPU. Hard-real-time control usually lives on an MCU (or an MCU core beside the MPU).
• Don't conflate the two MPUs. Microprocessor unit ≠ memory protection unit. Misreading this in a datasheet or interview is a classic slip.

TL;DR

• An embedded system is a computer dedicated to a specific function inside a larger product, usually with real-time constraints and fixed, limited resources.
• An MCU integrates CPU + flash + SRAM + peripherals on one chip, self-contained, instant-on, low-power, deterministic; runs bare-metal or an RTOS.
• An MPU (application processor) is essentially just a CPU needing external DRAM/storage; it runs a full OS, offers far more compute, but costs more power, money, and boot time.
• Choose MCU for real-time control / low power / low cost; choose MPU for an OS, rich UI, networking, or heavy compute, and many products use both.
• "MPU" means microprocessor unit or memory protection unit depending on context, don't confuse them.