Understanding CPU Architecture: From Basics to Modern Trends

CPU Basics

The Central Processing Unit (CPU) is the computational and control core of a computer. Its main functions are to interpret instructions and process data. A CPU consists of an arithmetic logic unit (ALU), a control unit, registers, and the buses that connect them.

Operation Cycle

The CPU works through four stages: Fetch, Decode, Execute, and Write‑back. It fetches an instruction from memory or cache into the instruction register, decodes it, and then executes it.

Key Performance Parameters

Important metrics include the clock frequency (often called the "base clock"), the external frequency, and the multiplier. The clock frequency equals the external frequency multiplied by the multiplier. Higher clock speeds generally mean faster data processing when core count and cache size are equal.

Why External Frequency and Multiplier?

CPU development outpaces many other hardware components, so the external frequency serves as a communication rate with the motherboard, while the multiplier allows the core to run at a higher effective speed.

Intel Turbo Boost

Intel Turbo Boost (also known as "Intel睿频加速技术") enables a processor to run above its nominal frequency when thermal and power headroom allow, allocating performance on demand.

Xeon Scalable Naming and Generations

In July 2017 Intel introduced the Purley platform and the new Xeon Scalable processors, replacing the E5/E7 naming with a tiered system: Platinum, Gold, Silver, and Bronze. The first digit indicates the tier, the second digit the generation, and the last two digits the SKU.

Tick‑Tock Development Model

Intel’s "Tick‑Tock" strategy alternates between a process‑technology update (Tick) and a micro‑architecture update (Tock) roughly every two years, reducing risk and keeping a steady performance cadence.

Multi‑Core and Hyper‑Threading

Multi‑core CPUs integrate multiple cores on a single die, allowing parallel execution of processes and reducing latency. Hyper‑Threading (simultaneous multithreading) presents each physical core as two logical cores, enabling the operating system to schedule more threads concurrently.

Heterogeneous and Many‑Core Computing

Beyond x86, heterogeneous units such as GPUs, FPGAs, and specialized accelerators are used for edge computing and AI workloads. "Many‑core" refers to CPUs with 32‑64 cores, while "multi‑core" typically denotes up to 10 cores.

CISC vs. RISC

CISC (e.g., x86) features complex, multi‑function instructions, offering high efficiency for specific tasks but increasing design complexity. RISC (e.g., ARM, Power, MIPS, RISC‑V) uses simpler instructions, yielding lower power consumption and higher per‑instruction throughput, though sometimes at the cost of raw single‑core performance.

ARM, RISC‑V, and MIPS Landscape

ARM enjoys a strong software ecosystem and is widely used in mobile and data‑center servers, with performance now comparable to mainstream x86. RISC‑V is an open ISA gaining traction. MIPS has seen limited recent development, though Chinese initiatives continue.

Chinese CPU Vendors and Server Ecosystem

Domestic Chinese companies such as Huawei (ARM), Phytium (ARM), Haiguang (x86), Loongson (MIPS), Zhaoxin (x86), and Shenwei (Alpha) are developing CPUs. Server manufacturers like Gigabyte, HPE, Lenovo, Ampere, Inspur, Great Wall, and others integrate these chips into their offerings.

Process Technology and Foundries

Performance is closely tied to manufacturing process and core design. Leading foundry TSMC provides advanced nodes down to 7 nm, while Chinese fabs such as SMIC, Hua Hong Semiconductor, and Huahong Microelectronics are progressing toward 14 nm and beyond.