Comprehensive Overview of System-on-Chip (SoC) Architecture, History, and Market Trends

System-on-Chip (SoC) is a large‑scale integrated circuit that extends a central processing unit (CPU) with audio‑video functions and dedicated interfaces, acting as the "brain" of intelligent devices. An SoC typically integrates application processors (AP), baseband processors (BP), memory, oscillators, power management, and a variety of peripheral interfaces such as USB, Ethernet, and ADC/DAC converters.

With the advancement of semiconductor processes, traditional micro‑controllers (MCU) can no longer meet the demands of modern smart terminals, leading to the emergence of SoCs that combine high performance, low power consumption, and flexibility on a single silicon die. SoCs are now pervasive in mobile computing, edge devices, IoT routers, and embedded systems.

Modern SoCs incorporate numerous functional blocks, including CPU, GPU, RAM, DSP, modems, high‑speed buses, and power‑management modules. Intellectual‑Property (IP) cores—validated, reusable design blocks such as CPU, GPU, DSP, VPU, and various interfaces—form the foundation of SoC design, and can be classified by function, implementation style (soft, firm, hard), or instruction‑set architecture.

Historically, ARM pioneered the IP‑core licensing model in 1990, enabling fabless vendors to integrate licensed cores into their SoCs. Over the years, companies like Motorola, Sony, and others have produced early SoC implementations based on licensed IP.

Typical SoC compositions include one or more processor cores (MCU, MPU, DSP, or custom cores), diverse memory types (RAM, ROM, EEPROM, flash), timing circuits (oscillators, PLLs), peripheral controllers, and standard I/O interfaces. Instruction‑set families span RISC architectures (ARM, MIPS, Power, RISC‑V) and CISC (x86), each dominating different market segments.

AI has become a critical driver for SoC development, with AI‑accelerators (Neural Engine, NPU, TPU, VPU) delivering substantial performance gains. Examples include Apple’s A14 SoC (5 nm, 16 % CPU boost, 10 % GPU boost, ~100 % AI accelerator improvement) and various ARM‑based server SoCs such as Ampere Altra, Amazon Graviton2, and Huawei Kunpeng.

Server‑class SoCs are divided into IA (x86/CISC) and RISC (ARM, MIPS, POWER) architectures. While x86 still dominates high‑performance computing, ARM’s power‑efficiency and ecosystem growth are reshaping the server market, especially in the post‑Moore era where AI, 5G, and big data increase cloud‑computing demands.

Supercomputing rankings (TOP500) show x86 as the prevailing architecture, yet ARM‑based A64FX SoCs (e.g., Japan’s Fugaku) have achieved top positions, highlighting the rising relevance of ARM in high‑performance domains.