// AN EMBEDDED HARDWARE GUIDE

BOARD
WORLD

From an IoT chip the size of a coin to the computer making driving decisions in an autonomous car — every board is really a point on the same spectrum. The difference is how much intelligence, how much precision, and how much power you actually need.

// POWER & COMPLEXITY SPECTRUM
ESP32Pocket-sized IoT chip
FRDM-K64FIndustrial real-time control
AURIX · S32 · RH850Automotive-grade MCUs
BeagleBone BlackLinux-powered mini computer
NVIDIA JetsonGPU-accelerated edge AI
DRIVE · Snapdragon RideAutonomous-driving computer
low power · simple tasks high power · artificial intelligence

Two Fundamental Worlds

The hundreds of products in the board world really lean on one of two philosophies. Knowing which one a board belongs to sets the first question you should ask when choosing hardware: does it run an operating system, or does it control the hardware directly?

NO OPERATING SYSTEM

Microcontroller (MCU)

CPU, RAM, and flash memory are fused onto a single chip. Code controls the hardware directly, with no operating system in between. The result: a predictable, deterministic response measured in milliseconds.

  • Instant boot
  • Deterministic timing
  • Low power draw
  • Limited memory and storage
RUNS LINUX

Single-Board Computer (SBC)

CPU, RAM, and storage are separate chips wired together on a board — the logic of a small PC. A full Linux runs on top, and programs reach the hardware through the operating system rather than touching it directly.

  • Boot takes seconds
  • Multitasking capable
  • Higher power draw
  • Generous memory and storage

Four Boards, Four Philosophies

Four different answers to the same questions: how much memory, how much intelligence, how much independence.

U1

ESP32

MCU · IoT
COREDual-core Xtensa LX6 · 240 MHz
MEMORY~520 KB SRAM (on-chip)
OSNone / FreeRTOS
STANDOUTBuilt-in Wi-Fi + Bluetooth
The default choice for IoT sensors and wireless prototypes.
U2

FRDM-K64F

MCU · Industrial
COREARM Cortex-M4 · 120 MHz
MEMORY~256 KB SRAM (on-chip)
OSNone / RTOS (Mbed)
STANDOUTWired, deterministic control
No Wi-Fi, but microsecond-level precision.
U3

BeagleBone Black

SBC · General Purpose
COREARM Cortex-A8 · 1 GHz
MEMORY512 MB DDR3 (separate chip)
OSFull Linux
STANDOUTHDMI + Ethernet output
A mini computer for robotics and software-heavy projects.
U4

NVIDIA Jetson

SBC · Edge AI
COREARM Cortex-A + CUDA GPU
MEMORY2–64 GB shared
OSLinux (JetPack)
STANDOUTParallel AI computation
BeagleBone Black's GPU-equipped cousin, built for AI.

Two Worlds, One Car

Hundreds of small, safety-certified MCUs manage the engine and brakes while one or two powerful AI computers process camera and lidar data. Two completely different board philosophies, running side by side in the same vehicle.

Classic ECUs · MCU class

AURIX · S32 · RH850

Infineon's AURIX, NXP's S32 family, and Renesas' RH850 — multi-core, ISO 26262 ASIL-D certified microcontrollers that run the engine, brakes, doors, and body electronics. An average car runs hundreds of these.

Autonomous driving · SBC+GPU class

NVIDIA DRIVE · Snapdragon Ride

Central computers that process camera, radar, and lidar data in real time to produce driving decisions, running Linux and AI frameworks. A car typically carries just one or two of these.

The industry is now consolidating most of those small ECUs not one by one, but into fewer, far more powerful "zone controllers" — built on higher-pin-count versions of the same AURIX, S32, and RH850 families.

The Full Spectrum

BoardClassOSPowerTypical Task
ESP32MCUNone / RTOSVery lowIoT, wireless sensors
FRDM-K64FMCUNone / RTOSLowIndustrial real-time control
AURIX / S32 / RH850Automotive MCUNone / AUTOSARLow–mediumEngine, brake, body control
BeagleBone BlackSBCLinuxMediumGeneral-purpose embedded Linux
NVIDIA JetsonSBC + GPULinuxMedium–highEdge AI, computer vision
DRIVE / Snapdragon RideAutonomous computeLinux + AIHighAutonomous driving decisions