news

"The CPU is undergoing an era of great historical change, from ordinary CPUs to AI CPUs, from a closed ecosystem to an open ecosystem, and from high-energy-consuming CPUs to high-performance CPUs." Sun Wenjian, founder and CEO of Cixin Technology, recently stated at its AI PC strategy and first chip launch conference that the era of edge-side generative AI has arrived.

2024 is known as the first year of the explosion of end-side AI. With the rapid development of artificial intelligence technology, the application of AI in terminal devices has gradually become popular, which has promoted the comprehensive upgrade of software and hardware architecture. Intel launched the Core Ultra processor, AMD released the Ryzen AI 300 series, Qualcomm demonstrated the Snapdragon X series platform, and NVIDIA plans to launch an AI PC chip based on the Blackwell core. Major terminal manufacturers have also joined the battle. Apple announced its artificial intelligence plan at WWDC 2024, enabling Siri to deeply connect multiple apps and provide users with a personalized smart assistant experience.

"Cisco is committed to promoting the development of the edge AI ecosystem with high-efficiency computing solutions, and making life, learning, and work smarter, more efficient, and more humane through the application of AI technology." Sun Wenjian, founder and CEO of Cisco, announced that after 15 months of research and development, 4 months of production, and 3 months of testing, the Cisco P1 designed specifically for AI PCs has fully met mass production requirements and will officially enter the productization stage.

With the in-depth deployment and application of generative AI big models in various industries, users have increasingly strong demands for energy efficiency, data privacy, and personalized customization. The hybrid deployment of cloud-based public big models and end-side private big models has become an industry consensus. As the best carrier of end-side generative AI, the PC industry has put forward new requirements for hardware, especially the AI ​​computing power of SoC as the core computing unit.

"The P1 chip, developed based on Arm technology, enables the advantages of the PC ecosystem based on the Arm platform to benefit a wider audience, bringing a high-performance, energy-efficient and AI-optimized platform that redefines the user experience for various workloads and use cases." James McNiven, vice president of product management at Arm's terminal business unit, said that AI is changing the way consumers use PCs.

Xu Yatao, executive vice president of sales and business at Arm Technology, also said, "In recent years, the two sides 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. In the future, we will continue to maintain close cooperation 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."

At present, the AI ​​capability of the edge is the key to realizing the global scale expansion of generative AI. The rapid development of edge AI forces the performance upgrade of edge hardware. The evolution of edge AI is closely related to the demand for generative AI. Software has first set off a wave of demand for performance improvement at the hardware level such as computing power and storage. Major chip leaders such as Qualcomm and Apple are competing to launch processor chips that can run generative AI models on the edge.

At the same time, new forms of smart terminals are constantly being implemented in terms of models, chips, operating systems, software applications, etc.: in terms of AIPC, starting from 2024, PC manufacturers will rush to the AIPC market. Canalys predicts that AI-compatible PCs are expected to reach a penetration rate of 60% in 2027; in terms of AI mobile phones, Huawei, Xiaomi, Honor, Samsung and other manufacturers continue to explore various aspects such as large models and operating systems; in addition, innovations in smart terminals such as AI smart wearables are also flourishing.

Cixin Technology has determined the development strategy of "one chip for multiple uses", targeting both the global and local markets, building an end-to-end AI ecosystem, focusing on the AI ​​PC field first, and creating a new generation of AI PC computing power base.

"Cixin Technology's new generation of AI PC computing power base is aimed at generative AI and supports hybrid artificial intelligence deployment." Sun Wenjianjin pointed out that by building a rich open ecosystem of software and hardware, it empowers developers and continues to explore the application scenarios and advantages of edge AI PCs.

Facing the changes and opportunities brought by the end-side generative AI to the PC industry, Cixin Technology focuses on the four major directions of boot firmware, kernel, graphics acceleration and AI solutions to carry out full-stack software innovation. "Hardware is like our body, and software gives the chip a soul," said Liu Gang, co-founder and vice president of software engineering of Cixin Technology.

Chu Ranzhou, co-founder and vice president of system engineering of Cixin Technology, was also at the press conference: "We are actively integrating into the PC industry chain and can support PC manufacturers to seamlessly switch from X86 CPU to Cixin P1 chip. We have already started cooperation with mainstream ODM manufacturers and well-known BIOS manufacturers in the industry."

Regarding the impact of AI PC on the domestic PC industry chain, Zhu Chen, deputy general manager of Kylin Software, said: "Generative AI has injected new vitality into the PC industry and brought new imagination space to the upper-level applications. For the first time, AI chips are integrated into the motherboard, and with the AI ​​subsystem at the operating system level, all upper-level applications are equipped with AI capabilities, bringing higher-quality development to the PC industry."

Liang Yejin, head of the end-side business of Alibaba Cloud Tongyi, said from the perspective of the coordination between inference engines and hardware: "In today's world where big models are flourishing, computing power on the end should not be wasted and should be fully utilized. Whether computing power can be used well depends on the coordination with the design of the upper-layer inference engine, which requires the joint efforts of the industry."

Regarding the matching problem between large models and hardware, Yao Hang, co-founder and COO of Qingmao Intelligence, said: "There are currently a large number of large models in China, and the hardware computing units of AI PCs are showing a trend of diversification. How to solve the matching of large models and hardware and the implementation of applications cannot be separated from the efforts of middle-layer partners. All software partners are required to coordinate the matching of models and hardware from the model side, framework side, inference engine side, and hardware side."