ARM Technology's heterogeneous computing power enables AI computing, and Cixin Technology's first AI PC chip is released

On July 30, the AI ​​PC strategy and first chip launch conference of Cixin Technology Group Co., Ltd. (hereinafter referred to as "Cixin Technology") was held in Shanghai, and its first heterogeneous high-efficiency chip product designed specifically for AI PCs, "Cixin P1", was officially launched. As the domestic new generation of AI PC computing power base, "Cixin P1" not only heterogeneously integrates Arm®v9 CPU core and Arm Immortalis™ GPU, but also carries self-developed business products such as "Zhouyi" NPU of Arm Technology (China) Co., Ltd. (hereinafter referred to as "Arm Technology"). With high-efficiency heterogeneous computing power resources, system-level security guarantees and strong technical ecological support, "Cixin P1" will better meet the application needs of generative AI in end-side scenarios such as PCs.

Figure 1: Cixin Technology releases its first AI PC chip

Xu Yatao, executive vice president of sales and business at Arm Technology, was invited to attend the press conference. While congratulating the official launch of "Chip P1", he also interacted with the guests on site on topics such as the development of the domestic PC industry and the construction of the Arm PC ecosystem.

Xu Yatao said: "In the field of personal computing, Arm architecture and generative AI are driving a new round of innovation in the entire PC industry. Cixin Technology is an important partner of Arm Technology, and is also a key participant and star company in the domestic Arm PC track. In recent years, the two parties have jointly promoted the iteration and upgrade of Arm CPU technology and led the expansion of the advanced Armv9 architecture in the PC application ecosystem. The launch of 'Cixin P1' is the fruit of the cooperation between the two parties, and it also injects new vitality into the Arm PC industry ecosystem. In the future, Arm Technology will continue to collaborate with Cixin Technology in multiple key areas such as product research and development and technology ecology, provide the industry with more reliable and efficient computing solutions, and promote the widespread application of terminal-side AI in various fields."

Heterogeneous computing power combination breaks the challenge of generative AI computing power

Currently, with the continued rise of the AI ​​big model wave, PC, as one of the important carriers of generative AI, has become a key industrial highland for current giants to compete for. However, while generative AI quickly empowers hardware, it also puts forward higher requirements on its computing power, energy efficiency, and software and hardware collaboration. For this reason, heterogeneous computing combining various computing power units has become the key to breaking the deadlock.

Figure 2: Sun Wenjian, founder and CEO of Cixin Technology, released the first heterogeneous high-efficiency SoC Cixin P1

In order to cope with the new computing power and performance challenges brought by generative AI, "Cixin P1" adopts the industry-leading 6nm manufacturing process, and through the design of multi-core heterogeneous and dedicated NPU, it integrates Arm Cortex®-A720 CPU, Arm Immortalis-G720 GPU and Arm Technology's "Zhouyi" NPU and other self-developed business products, providing rich AI heterogeneous computing resources, comprehensive security engine protection, and more outstanding energy efficiency performance for different scenarios.

Specifically, the NPU designed based on the third-generation "Zhouyi" architecture, as a key computing power resource for edge AI applications, will provide core potential for the distributed landing evolution of large models. Through targeted optimization of low power consumption and high computing power, the "Zhouyi" NPU is not only capable of long-term, high-load task processing, but also supports multi-core and multi-Cluster computing power expansion. In the design of the "Zhouyi" NPU, Arm Technology has deeply optimized various image and video AI model algorithms from multiple dimensions such as performance, accuracy, bandwidth, scheduling management, and operator support, and has completed the adaptation of mainstream large models, further enhancing its performance in the face of high-performance AI computing needs.

In addition, "Cixin P1" has outstanding performance in general computing, graphics image acceleration, rendering, etc. through heterogeneous integration of Arm Cortex-A720 CPU and Arm Immortalis-G720 GPU. Among them, the Arm Cortex-A720 CPU, as the first 64-bit ultra-efficient CPU based on the Armv9.2 architecture, can bring breakthrough performance within a limited power consumption range and can be optimized for various computing use cases. It is particularly suitable for consumer technology computing fields for security and machine learning. The Arm Immortalis-G720 GPU is the first flagship product to adopt Arm's new fifth-generation GPU architecture. By introducing deferred vertex shading (DVS) technology, it effectively saves power consumption and improves frame rate, thereby achieving higher quality graphics performance.

At the software development level, the "P1 Core" will also introduce the Arm Kleidi software library for popular AI frameworks, aiming to enable developers to easily and conveniently obtain the excellent AI capabilities of Arm CPUs on a variety of devices. Currently, most AI inference workloads from the cloud to the edge are running on these Arm CPUs worldwide.

Innovation leads, cooperation leads, and Arm PC ecosystem prospers

In order to better promote the construction of the AI ​​PC industry chain, a special AI PC industry chain strategic cooperation launch ceremony was held during the conference. Arm Technology, together with many partners such as Cixin Technology, Lenovo Group, Tongfang Dingxin, Wanmos, Tongxin Software, Kylin Software, Jiangbolong, Baiao Software, Wuwen Xinqiong, etc., will give full play to their respective advantages in IP design, full-stack software development and system design, accelerate the commercialization of domestic AI PC innovative products, and promote the sustainable development of the industry.

Figure 3: AI PC industry chain strategic cooperation launch ceremony

In addition, at the roundtable forum of the press conference, Xie Wei, general manager of ARM Global Service Marketing, and guests from Cixin Technology, Lenovo Group, Kylin Software, Alibaba Cloud, Qingmao Intelligence and other companies conducted in-depth discussions and exchanges on core topics such as the hardware requirements for edge AI reasoning and the development of hybrid AI, the impact of generative AI on the PC software stack, and the opportunities and challenges of the AI ​​PC industry.

Figure 4: Roundtable Forum

After more than 30 years of innovation and deep cultivation, Arm technology has been widely used in chips in almost all fields, becoming the world's most widely used computing platform. The shipments of chips based on Arm architecture have exceeded 280 billion to date, covering diverse fields such as smartphones, PCs, smart cars, cloud data centers, etc.

Today, with the continuous expansion of AI big models on the edge, Arm Technology, as a core enterprise in the upstream of the chip industry chain, continues to deepen local innovation practices on the basis of absorbing and leveraging Arm technology and ecological advantages, and has independently developed and launched NPU, CPU, information security, multimedia and other product matrices. Adhering to a systematic and overall thinking, Arm Technology actively promotes AI architecture upgrades, hardware implementation and software optimization, providing the industry with solid and reliable underlying computing power support, and accelerating the implementation and application of edge AI in many industries.