The first change to the C++ sorting library: artificial intelligence improves computer programming language [with analysis of mainstream products in the artificial intelligence chip industry]

The latest news points out that the artificial intelligence "Alpha Development" (AlphaDev) created by the British Deep Thinking Company has successfully discovered and improved computer sorting algorithms. These improved algorithms have been integrated into the widely used C sorting library. This groundbreaking result is published in the latest issue of Nature and is the first major update to the C library in more than a decade. This breakthrough progress benefits from the powerful artificial intelligence processing capabilities and deep thinking of "Alpha Development", demonstrating the potential and application prospects of artificial intelligence technology in the computer field.

GPU, FPGA, and ASIC have become the mainstream of the artificial intelligence chip industry

Currently, artificial intelligence chips can be divided into CPU, GPU, FPGA, ASIC, and brain-inspired chips according to their technical architecture; according to their position in the network, they can be divided into cloud AI chips, edge AI chips, and terminal AI chips; according to Its goals in practice can be divided into training chips and inference chips. With its powerful algorithm carrying capacity and ultra-high processing speed, artificial intelligence chips are widely used in a variety of scenarios, such as intelligent face recognition or intelligent voice recognition, server big data analysis for processing ultra-high databases, and processing of changing traffic information and Various types of sensor information in the field of autonomous driving, as well as the intelligence of robots, etc.

China’s AI chip market continues to be hot, with GPU market accounting for the largest share

China’s artificial intelligence chips will continue to become the focus of development in the next few years. GPU, FGPA, and ASIC are the mainstream development of China’s artificial intelligence chip industry. Among them, GPU was originally designed to accelerate the rendering of 3D graphics. With the development of the digital economy, With continuous development, GPUs have become more flexible and programmable, allowing developers to use the powerful functions of GPUs to significantly accelerate additional workloads in areas such as high-performance computing and deep learning. According to data disclosed by IDC, in the first half of 2021 in China's artificial intelligence chip industry, GPU has significantly become the first choice for data center acceleration, accounting for more than 90% of the market share, while non-GPU chips such as ASIC, FPGA, and NPU have a relatively small market share. Less, the overall market share is close to 10%.

GPU market size continues to rise

The domestic artificial intelligence chip market demand continues to expand, among which GPUs are mainly scientific computing types. They have an advantage in the artificial intelligence chip market with their strong computing power and mature product performance. According to estimates, China's GPU server market size will be 28.58 billion yuan in 2021; it is expected that China's GPU server market size will continue to rise in the future from 2023 to 2025.

The performance gap between domestic GPUs and the international leading level is accelerating. The performance of JM9 series graphics processing chips is close to NVIDIA GeForce GTX 1050 in terms of graphics rendering GPU. Although there is a huge gap between Jingjiawei's products and international cutting-edge GPUs, as a GPU that is completely independently developed by a Chinese company, adopts forward design, and has independent intellectual property rights, it has made great progress.

China Post Securities believes that it is time for domestic ASIC substitution and FPGA development opportunities are great. The ASIC market has not formed a monopoly, and domestic manufacturers still have considerable opportunities to seize the market. Compared with other AI chips, the development threshold of FPGA chips is higher and requires a large investment of resources, so there are certain barriers. Chinese FPGA manufacturers are still in their infancy, but with the huge demand brought by 5G and AI industrialization, domestic FPGA enterprise technology is expected to continue to achieve breakthroughs and usher in rapid growth.

Foresight Economist APP Information Group

The above data refers to the Qianzhan Industry Research Institute.

At the same time, Qianzhan Industry Research Institute also provides solutions such as consulting and consultation.

More in-depth industry analysis is available in [Forward-looking Economist APP], and you can also communicate and interact with 500 economists/senior industry researchers.

The above is the detailed content of The first change to the C++ sorting library: artificial intelligence improves computer programming language [with analysis of mainstream products in the artificial intelligence chip industry]. For more information, please follow other related articles on the PHP Chinese website!