How Zhaoxin’s KX-7000 and KH-40000 CPUs Challenge Global Chip Leaders

Background

In the era of rapid information‑technology advancement, chips are the core components that determine both industrial competitiveness and national security. Zhaoxin, a domestic Chinese chip maker, holds a permanent x86 instruction‑set license and has been steadily advancing its own processor designs.

Historical Evolution of Zhaoxin Processors

Early Exploration (2013‑2016)

Zhaoxin was founded in 2013 as a joint venture between Shanghai’s state‑owned Lianhe Investment and Taiwan’s VIA Technologies, acquiring a permanent x86 license. The first ZX‑C series, built on a 28 nm process, delivered desktop‑class SoCs and laid the foundation for a domestic x86 ecosystem, despite performance gaps with international rivals.

Iterative Improvements (2017‑2020)

KX‑5000 (2017): 28 nm, 2.0 GHz, dual‑channel DDR4, PCIe 3.0, USB 3.1 Gen2.

KX‑6000 (2019): 16 nm, 25 % IPC uplift, up to 3.0 GHz, DDR4 support, narrowing the performance gap with global competitors.

Breakthrough Architecture (2021‑Present)

The KX‑7000 series (2023) introduces the self‑designed “Yongfeng” micro‑architecture on a 12 nm process, offering up to 8 cores/16 threads, 3.7 GHz boost, DDR5, PCIe 4.0, and an integrated graphics engine that doubles compute performance and quadruples graphics capability. The KH‑40000 server line (2023) follows a similar trajectory with 7 nm, up to 32 cores/64 threads, 120‑180 W TDP, 8‑channel DDR4‑3200 (up to 2 TB), and NUMA‑aware multi‑socket support.

Technical Highlights

Self‑Designed Micro‑Architecture

Zhaoxin’s proprietary cores—named Lujiazui, Yongfeng, and Shiji‑Dadao—feature deeper pipelines, larger L2/L3 caches, and refined branch prediction, delivering higher instruction‑per‑cycle efficiency and better multi‑thread scaling.

High‑Speed I/O

Support for DDR5 memory and PCIe 4.0 dramatically increases bandwidth while reducing power consumption, enabling faster data exchange for GPUs, NVMe SSDs, and other accelerators.

Security Integration

Custom extensions to the x86 ISA include hardware acceleration for Chinese national‑cryptography (SM) algorithms and a secure‑compute framework, enhancing data protection for government and financial workloads.

Core Technology Deep‑Dive

x86 License Extension

Leveraging the permanent x86 license, Zhaoxin adds proprietary instructions—especially for SM‑based cryptography—while maintaining full software compatibility with the global x86 ecosystem.

Chiplet Interconnect

The KX‑7000 adopts a chiplet architecture that stitches together heterogeneous dies (CPU core, I/O, cache) via high‑speed interconnects, reducing design risk, enabling mixed‑process integration, and improving overall performance.

High‑Performance IP Design

Zhaoxin designs its own high‑speed I/O IP blocks (PCIe, DDR, USB) that match international specifications, ensuring tight integration with the core and reducing latency.

Application Scenarios

Trusted Computing (Xinchuang) : Lenovo Kaitu and N8 Pro laptops use KX‑6640MA and KX‑7000 CPUs with DDR5 and PCIe 4.0, targeting government procurement and secure desktop workloads.

Education : Over 500,000 units deployed across 31 Chinese provinces, including AI‑enabled AIPC devices that run local large‑language models on Zhaoxin CPUs combined with custom NPU accelerators.

Enterprise : Inspur Yinxin CN3160 servers equipped with dual KH‑40000 CPUs deliver robust performance for cloud, virtualization, and database workloads in data centers.

Conclusion

Zhaoxin has made significant strides in the x86 domain, progressing from modest 28 nm SoCs to high‑performance 7‑12 nm multi‑core processors with advanced I/O, security, and chiplet technologies. While the company still faces intense competition from established global vendors, continued R&D investment and ecosystem building are essential for sustaining its growth and contributing to China’s semiconductor independence.