7 nanometers! Huawei releases the latest chip called Kunpeng, with 64 cores and unmatched performance

Xu Wenwei, Huawei Director and President of Strategic Marketing, said:

The industry's highest performance ARM-based processor

30% reduction in power consumption at the same performance

Kunpeng 920 is the industry's highest-performance ARM-based processor. The processor uses a 7nm manufacturing process, is based on ARM architecture authorization, and is independently designed by Huawei. Through a series of micro-architecture designs such as optimizing branch prediction algorithms, increasing the number of computing units, and improving the memory subsystem architecture, the processor performance is greatly improved. At a typical main frequency, the SPECint Benchmark score exceeds 930, which is 25% higher than the industry benchmark. At the same time, the energy efficiency ratio is 30% better than the industry benchmark. Kunpeng 920 provides stronger performance for data centers with lower power consumption.

Kunpeng 920 can run at up to 2.6GHz, and a single chip can support 64 cores. The chip integrates 8-channel DDR4, and the memory bandwidth exceeds the industry mainstream by 46%. The chip integrates 100G RoCE Ethernet card function, which greatly improves the system integration. Kunpeng 920 supports PCIe4.0 and CCIX interfaces, providing a total bandwidth of 640Gbps. The single-slot interface rate is twice the industry mainstream rate, effectively improving the performance of storage and various accelerators.

Huawei TaiShan

The industry's best-performing ARM server

Huawei also launched the TaiShan series of server products based on Kunpeng 920, including three models: balanced, storage, and high-density.

TaiShan series servers are mainly aimed at scenarios such as big data, distributed storage, and ARM native applications. They leverage the advantages of ARM architecture in multi-core and high energy efficiency to build a new high-performance, low-power computing platform for enterprises. For example, in big data scenarios, they achieve multi-core high concurrency and resource scheduling and optimization, and improve computing performance by 20%. Based on TaiShan servers, Huawei Cloud will also provide elastic cloud services, bare metal services, and cloud phone services.

We will work on three levels to actively build an open, cooperative and win-win ARM ecosystem

Huawei continues to promote industry cooperation at the three levels of hardware, basic software, and applications. For a long time, Huawei has worked with industry organizations such as GCC, Linaro, and OEHI, as well as partners such as Hortonworks, Microsoft, Oracle, SAP, SUSE, Ubuntu, and Zhongbiao Software to create an open, cooperative, and win-win ecological environment. At the hardware level, Huawei has become a core member of Linaro; at the basic software level, Huawei has become a platinum member of OpenStack and a Founder member of CNCF; at the application level, Huawei participates in the Green Computing Industry Alliance GCC, which has released the "Green Computing Industry Alliance Server Standard White Paper" and built a green computing open source community; Huawei participates in the "Open Edge + HPC Initiative" in Europe.

Huawei believes that the intelligent society where everything is connected, sensed and intelligent is accelerating. ARM-based smart terminal applications are developing rapidly and cloud collaboration is emerging. At the same time, new services under cloud computing are making data types more diverse, such as big data applications, distributed storage and some edge computing. These scenario applications have clear requirements for multi-core (Many Core) high-efficiency computing, and ARM computing systems with advantages in performance and power consumption will play a role.

Judging from industry trends and application needs, the era of diverse computing is coming. Various data types and scenarios drive the optimization of computing architecture. The combination of multiple computing architectures is an inevitable choice to achieve optimal performance computing.

Xu Wenwei finally said:

Content Statement: The content of this article comes from Huawei, and the copyright belongs to the author. Any views in this article are for discussion purposes only, and do not constitute any investment advice, nor do they represent the position of this public account. Users who invest based on this article and any other views of this public account must bear their own risks and responsibilities. This public account does not assume any responsibility for any consequences caused by this.