SoC

System on Chip (SoC)

A System on Chip (SoC) is an integrated circuit that combines all major components of a complete electronic system — processor cores, memory, peripherals, communication interfaces, and often analog or RF circuits — onto a single piece of silicon. SoCs are the dominant architecture in 2026 for smartphones, IoT devices, automotive ECUs, and edge AI products because they minimize board area, power consumption, and BOM cost while maximizing data-path performance through on-chip communication. The smartphone market alone deploys over 1.4 billion SoCs annually.

Key Facts

Aspect	Detail
Definition	Complete computing system integrated onto one silicon die
Core components	CPU(s), memory controllers, GPU, accelerators, peripherals, I/O, often radio
Process nodes (2026)	Smartphones: 3–5nm. Automotive: 7–28nm. IoT/embedded: 22–55nm
Typical complexity	10 billion – 50 billion transistors (advanced)
Time to develop	18–36 months for a new custom SoC; faster for derivatives
NRE cost	€5M – €100M+ (full-custom on advanced nodes)
Standard ARM-based SoC examples	Apple A18, Qualcomm Snapdragon 8 Gen 4, MediaTek Dimensity 9400

SoC vs Microcontroller vs FPGA

Criterion	SoC	Microcontroller (MCU)	FPGA
Complexity	Highest (multi-core, GPU, AI, RF)	Lowest (single core + peripherals)	Medium-high (configurable logic + IP)
Software stack	Linux, Android, Windows	Bare-metal or RTOS	Custom logic + soft processors
Typical use	Smartphones, set-top boxes, edge AI	Sensors, actuators, IoT endpoints	Signal processing, radar, prototyping
Power	1–15 W typical	1–100 mW	1–30 W
Boot time	Seconds (Linux)	Milliseconds	Hundreds of milliseconds (bitstream load)

For a deeper MCU vs MPU comparison, try our interactive MCU vs MPU Advisor tool.

SoC FPGA — The Best of Both Worlds

An SoC FPGA integrates a hard processor system (ARM Cortex-A53/A72 or RISC-V) with programmable FPGA logic on a single die. This hybrid is increasingly the architecture of choice for advanced embedded systems because it combines:

General-purpose computing — Linux on the hard CPU cores for networking, GUI, cloud connectivity, machine learning frameworks
Hardware acceleration — FPGA fabric for custom signal processing, deterministic control, custom neural network inference, sensor fusion at line rate
Tight coupling — Multi-hundred-Gb/s on-chip bandwidth between CPU and FPGA fabric (AXI bus)
I/O flexibility — High-speed serial transceivers (PCIe, 10G/25G Ethernet, MIPI) shared between CPU and FPGA logic

Leading SoC FPGA families in 2026: AMD Zynq UltraScale+ MPSoC and AMD Versal Adaptive SoC (ARM Cortex-A72 + FPGA + AI Engines), Intel Agilex SoC (ARM Cortex-A53 quad-core + FPGA), Microchip PolarFire SoC (RISC-V cores + FPGA, ideal for security-critical and rad-tolerant applications), and the EU-sovereign NanoXplore NG-Ultra (ARM Cortex-R52 + space-grade FPGA fabric).

Common SoC Categories

Mobile / consumer SoCs — Apple A-series, Qualcomm Snapdragon, MediaTek; multi-cluster CPU + integrated GPU + NPU + 5G modem
Automotive SoCs — NXP S32, Renesas R-Car, NVIDIA Drive Orin; ISO 26262 functional safety, AI accelerators, multi-camera input
Networking SoCs — Broadcom, Marvell, Microchip switching silicon; multi-Tbps packet processing
IoT SoCs — Espressif ESP32-S3/C6, Nordic nRF52/53, Realtek; Wi-Fi/BLE/Thread radios integrated with MCU core
AI inference SoCs — Hailo-8/10, NVIDIA Jetson Orin, Google Edge TPU; optimized for on-device neural networks
SoC FPGAs — AMD Zynq, Intel Agilex SoC, Microchip PolarFire SoC; reconfigurable hardware acceleration

When to Choose an SoC Approach

Choose an SoC architecture when your product needs significant compute (Linux-class), multi-protocol connectivity (Wi-Fi/BLE/Cellular), and graphics or video processing — the integration savings vs. discrete components are substantial in BOM, board area, power, and EMC compliance. For lower-complexity embedded products with single-task firmware, a microcontroller is more cost-effective. For low-volume products needing custom signal processing or reconfigurability, FPGA or SoC FPGA wins.

CRA Compliance and SoCs

Under the EU Cyber Resilience Act, products built around SoCs must meet baseline cybersecurity requirements — secure boot anchored in silicon, authenticated firmware updates, hardware-protected key storage. Modern SoCs include security building blocks (TrustZone on ARM, secure enclaves, eFuses, hardware random number generators) that simplify CRA conformity assessment. SoC vendor choice has direct CRA implications: SoCs with PSA Certified Level 2 or SESIP3 certification dramatically reduce the engineering work needed to demonstrate compliance.

FPGA — Reconfigurable logic, often combined with SoC as SoC FPGA
ASIC — Modern SoCs are ASICs at high volume; SoC is the design pattern, ASIC is the manufacturing reality
MCU vs MPU — Microcontrollers and microprocessors are the simpler cousins of full SoCs
CRA — EU Cyber Resilience Act, which imposes security requirements on SoC-based products
Edge AI — Modern edge AI typically deploys on SoCs with integrated NPU or FPGA acceleration

Inovasense SoC Capabilities

Inovasense specializes in SoC FPGA designs for Linux-class embedded systems requiring real-time hardware acceleration. We integrate AMD Zynq, Intel Agilex SoC, Microchip PolarFire SoC, and NanoXplore for European defense and dual-use applications. Our embedded systems development team builds the full software stack — bootloader, secure firmware update, Linux board support package, real-time hardware drivers — alongside the silicon selection and PCB design. For products targeting EU market entry, we manage the CRA conformity assessment for SoC-based connected products.

Official References

Regulation (EU) 2024/2847 (CRA) — EU Cyber Resilience Act, scope includes SoC-based products with digital elements
ARM PSA Certified Framework — Industry-standard security certification framework for ARM-based SoCs
ISO/SAE 21434 — Cybersecurity engineering standard for road vehicle SoCs