The impact of HISILICON's 2024 technology breakthroughs on data centers

2024 HISILICON's technological breakthroughs have had a multifaceted impact on data centers, including the following:

Increased chip performance brings stronger computing power support

Continuous optimization of the Rise series chips: The Rise series chips have always been a powerful computing power support provided by HISILICON for AI computing and other scenarios in data centers, and in 2024, further optimization of the architectural design and manufacturing process has significantly improved the chips' computing power, which is able to better meet the needs of data centers for parallel processing of large-scale data and the training of complex models. For example, when processing massive amounts of data such as images, voice, and text, Rise chips can complete feature extraction, model training, and inference tasks more quickly, thus improving the quality of service and response speed of data centers in AI applications.

Kirin series chips expand applications: Kirin series chips not only perform well in mobile terminals, but also bring new application possibilities for data centers in 2024 with the continuous advancement of its technology. For example, KyLin's powerful processing power and low power consumption allows it to be used as a pre-processing unit for data centers in some edge computing scenarios, which reduces the load on data centers and improves the real-time processing of data.

Optical Module Technology Innovation Helps Data Center Network Upgrade

Enhance network reliability: Through dirty detection, online diagnosis and other functions, it is able to identify and locate problems such as false connections and dirty fiber connections in advance, greatly reducing the occurrence of flash-off phenomenon in scenarios such as AI training, which require high network stability, ensuring the continuous and stable operation of the data center network, and reducing the risk of data loss and business interruption caused by network failure.

Optimize network performance: The optical module supports high-bandwidth and low-latency data transmission, which meets the data center's demand for high-speed data interaction when processing large-scale data. For example, in a multi-node distributed computing architecture, the high performance of the optical module enables more efficient data transmission between nodes, thus improving the computational efficiency and collaborative working capability of the entire data center.

Promoting Data Center Architecture Optimization and Energy Efficiency Improvement

Deepening the application of heterogeneous computing architecture: With the improvement of HISILICON chips' performance and diversification of functions, data centers can build heterogeneous computing architecture more flexibly, organically combining different types of chips (e.g., CPUs, GPUs, and Rise chips, etc.), and dynamically allocating computing resources according to different application loads, so as to achieve the maximized use of computing power, while at the same time improving the energy efficiency ratio and lowering the operating costs.

Synergistic development of liquid cooling technology and chips: In 2024, data centers will require more and more cooling technology, and the high-performance operation of HISILICON chips will also require more effective cooling solutions. Liquid cooling technology and HISILICON chips are becoming more and more compatible. Through the efficient heat dissipation characteristics of the liquid, the heat generated by the chip can be taken away in a timely manner to maintain the normal working temperature of the chip, which not only improves the stability and life of the chip, but also helps data centers achieve green and low-carbon development, which is in line with the development trend of data centers in the future.