ai drives the growth of cpu demand, and domestic mainstream manufacturers have huge room for growth
2024-10-01
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
in the widely discussed ai computing power base, whether it is the "cpu+gpu and other accelerator cards" on the cloud or the "cpu+gpu+npu" on the device side, the cpu is indispensable.
cpu, or central processing unit, is often composed of a control unit, an arithmetic logic unit (alu) and registers. it is responsible for reading, decoding and executing instructions. it has high requirements on r&d technology and ecological construction, and has high requirements on computer performance and operating efficiency. have important influence.
as the computing and control center of computers, under the wave of ai development, my country's computer cpu industry has entered the "fast lane": the construction of intelligent computing centers has boosted demand for servers, and the upgrading of aipcs is also driving demand growth.
compared with gpu, the development of domestic cpu is more mature. according to data from the toubao research institute, china’s cpu localization rate will be 30%-40% in 2022. at the same time, china’s cpu localization rate is still increasing. in the recent centralized procurement by operators, the proportion of domestically produced cpus is close to 70%.
the explosion of ai has boosted cpu demand
with the rapid development of large models, the demand for intelligent computing power has exploded, and the growth in ai server shipments has driven the demand for cloud cpus. as the core component of the server, the cpu's performance and power consumption directly affect the overall performance of the server.
techinsights data shows that by 2028, the global server market will reach us$273 billion, with a compound annual growth rate of 18%. servers designed for ai applications will account for 59% of the market share, with a compound annual growth rate of 49%. in terms of the chinese market, idc predicts that china's ai server market will reach us$9.1 billion in 2023, a year-on-year growth of 82.5%; it will reach us$13.4 billion in 2027, with an average annual compound growth rate of 21.8%.
in the pc market, 2024 is recognized as the first year of aipc, and aipc is expected to drive a new wave of replacements. the industry believes that the current development of aipc is similar to the development of early windows systems, and it is expected to replicate the success of early windows systems in the future.
canalys predicts that global pc shipments in 2023 will be 249 million units, a year-on-year decrease of 12.4%. pc shipments in 2024 are expected to be 267 million units, a year-on-year increase of 7.6%. aipc shipments in 2024 will be around 20 million, and in 2027 60% of computers will have ai processing capabilities, and shipments will exceed 175 million units in 2027.
driven by ai servers and aipc, cpu demand will also grow rapidly. according to data from the toubao research institute, china's computer cpu market size has grown steadily and rapidly at an average growth rate of 6.91% from 2019 to 2021, and is expected to reach 510.949 billion yuan by 2027.
the ai wave set off by large models not only drives the demand for cpus to grow, but also puts forward higher requirements for cpu functions.
"many ai applications can actually be calculated using different xpu processing units. cpu programming is more flexible, and operator acceleration can be processed through a general compiler. it is more versatile and flexible than npu." zhou xixin, general manager of ecological strategy of xixin technology jie told china business news, "for example, our armv9 cpu includes vector extension instruction acceleration, which can better help run large language models on the device side. the latest iphone 16 series has also upgraded the cpu, including matrix operation instruction acceleration, to better support those cpu-based apple intelligence applications."
zhou jie said that in the future, this core technology will continue to innovate in cpu products, provide more powerful vector and matrix processing capabilities, and combine the heterogeneous processing capabilities of gpu and npu to more efficiently meet the growing needs of large language models and multi-modal large-scale languages. model and other end-side ai computing power requirements.
the battle over instruction set architecture
chip architecture is the basis for designing and building microprocessors or other integrated circuits, and defines key characteristics such as the chip's organizational structure, functional units, data flow, and instruction set. the instruction set architecture is the top-level design in the chip architecture. it defines the set of all instructions that can be executed by computer hardware. it is the core part of computer system design and has an important impact on both upper-layer software and lower-layer hardware.
according to different design principles, instruction set architectures are mainly divided into two camps: complex instruction set computing (cisc) and reduced instruction set computing (risc), from which instruction set architectures such as x86, arm, and risc-v are derived.
x86 was launched by intel and is a cisc design. it has a rich instruction set and supports multiple addressing modes and complex operations. currently, processors equipped with the x86 architecture instruction set are mainly used in pcs and servers. since intel strictly controls patent licensing, the x86 market is basically monopolized by intel.
in the server market, x86 is the mainstream architecture for cpus. according to idc data, the x86 architecture server market share will be 88% in 2023, and china’s x86 server market shipments will be 3.62 million units, and it is expected to grow by 5.7% in 2024.
the arm architecture is based on risc design. the arm core microarchitecture ip has a variety of choices, the design is streamlined and reliable, and it performs well in the field of low power consumption. arm's business model focuses on licensing the processor core architecture rather than directly designing and manufacturing chips. therefore, its neutral status allows arm to widely promote the ecosystem through licensing. it is reported that the arm architecture is currently widely used in the field of mobile intelligence represented by mobile terminal chips (mobile phones, tablets, etc.), set-top boxes, video surveillance, etc.
overall, cloud servers currently mostly use the x86 architecture, and the edge side uses the arm architecture. however, the landscape has also been changing in recent years, and the arm architecture has gradually penetrated into the server field. "now many internet manufacturers around the world have begun to use arm architecture to design their own server chips, such as amazon, google, microsoft, etc., as well as domestic alibaba." zhou jie told china business news. in addition, aipc is equipped with npu, and arm architecture may have more advantages. arm ceo rene haas said at computex 2024 that arm expects to capture more than 50% of the windows pc market within five years.
the open reduced instruction set architecture risc-v has also begun to rise in recent years. risc-v is completely open source and adopts the loose bsd protocol. enterprises can use it completely freely and for free. it also allows enterprises to add their own instruction sets without having to share them openly to support differentiated development. at the same time, the architecture is simple, short and concise, and different parts can be used. it can be composed together in a modular manner to meet various application scenarios through a unified architecture. it is understood that many domestic start-ups are currently starting to use risc-v for chip design. shd group predicts that soc revenue based on risc-v is expected to reach us$92 billion in 2030, with a compound annual growth rate of 47% from 2021 to 2030.
"the risc-v architecture is currently more used in some fixed scenarios, such as iot devices, but its support for general ecology, such as pc and server fields, requires more resource investment and r&d support." zhou jie told china business news .
in addition, some companies have developed their own instruction sets. loongson zhongke (688047.sh) announced in 2023 that it has independently developed the loong arch architecture with completely independent property rights based on the mips architecture. sunway has developed the sw64 architecture based on the alpha architecture.
"ecology is a very important issue. the x86 ecology is undoubtedly the most mature, and the arm architecture ecology is becoming more and more mature." zhou jie said, "compared with the x86 and arm architecture ecology, the ecology of the self-developed architecture will have certain advantages. gap. if you need to support a more universal ecosystem, the entire industry chain will require a lot of investment. for example, each software manufacturer will need to adapt to the self-developed architecture and perform related software optimization.”
comparison of six major mainstream manufacturers
looking at the market structure, intel and amd currently lead the general cpu market.
according to a report from cpu market tracking agency mercury research, in the second quarter of 2024, intel accounted for 78.9% of the market share in the pc market, and amd accounted for 21.1%; in the server cpu market, intel accounted for 75.9% of the market share, and amd accounted for 21.1%. is 24.1%.
in the domestic market, domestic cpu manufacturers are divided into three major schools: one is self-developed, such as loongson and sunway mentioned above; the other is arm, such as feiteng, a subsidiary of china great wall (000066.sz), and huawei hisilicon both use arm architecture; the third is x86. shanghai zhaoxin and haiguang information (688041.sh) both use the x86 architecture.
as xinchuang continues to advance, domestic cpus may gain greater room for growth. canalys data shows that pc purchases by the chinese government and education departments will reach 2.72 million units in 2023, accounting for 6% of national shipments. in march 2024, the central government procurement network issued an announcement stating that when purchasing computers, the safety and reliability evaluation requirements of the cpu and operating system should be included in the procurement requirements. canalys believes that the above six companies are all brands that meet the conditions for domestic cpus in the notice.
"the industry is working hard to promote the development of advanced manufacturing processes and upstream eda tools, which have a great impact on the competitiveness of chip products." zhou jie said, "based on the current domestic industry status, it is not a big problem to design a chip. the key is how to produce it after it is designed, and how to build a good ecosystem to help developers and end users use it better. this is particularly important and reflects the company's core capabilities. "
(this article comes from china business news)
