Shulou(Shulou.com)11/24 Report--

Today, Chen Baoli, vice president and general manager of Intel data Center and artificial Intelligence Group, delivered a speech entitled "Intel AI Strategy for the Chinese Market" at the China AIGC Innovation and Development Forum at the China International Trade in Services Fair in 2023. Starting from the two dimensions of general computing and accelerated computing, Chen Baoli expounds in detail how Intel continues to explore and innovate with a wide range of Chinese ecological partners through a comprehensive heterogeneous product portfolio and an open software stack, so as to seize the opportunities of the AI era and create unlimited possibilities.

Chen Baoli pointed out, "committed to creating world-changing technologies, Intel has been creating higher value for Chinese customers and promoting deeper local innovation and development through continuous technological changes since it entered China in 1985. In the early days of the AI' Golden Age, with its deep accumulation in chip manufacturing and innovation, Intel is also taking advantage of its industry leadership to support the sustainable development of the industry through a comprehensive layout in cloud, networking, edge, client, and other areas. "

The data flow of AI includes a wide range of complex workloads and multimodal data sets. In the face of the computing requirements of AI, there is not a general solution. Today, due to multiple technical and economic factors, a large proportion of AI workloads run on general-purpose processors. The large-scale popularity of x86 architecture and its built-in AI capabilities make general-purpose processors an ideal solution for AI data streams.

Based on this, Intel built the fourth generation Intel ®strong ®scalable processor with built-in Intel ®AMX acceleration engine in enterprise hardware, which provides up to 10 times the performance improvement in AI reasoning and training than the previous generation, and supports most large-scale AI models, including real-time, medium throughput, low-latency sparse reasoning, as well as small and medium-sized training and edge reasoning. In addition, it makes it easier for customers to deploy Intel solutions through a broad ecosystem, software tools designed to simplify processes, and optimized compilers, as well as continued investment in higher-level software stacks such as oneAPI and OpenVINO, making it easier for developers to use the AI framework they are familiar with.

Based on the fourth-generation Intel ®®scalable processor integrated with Intel ®AMX, Baidu has also built its own artificial intelligence natural language processing model ERNIE-Tiny. The lightweight version launched by Baidu on the core product of Flying Propeller Wenxin model has been improved many times after the introduction of Intel ®AMX, as well as satisfactory reasoning performance on the general CPU platform, which can further expand the scope and help more users to deploy Wenxin model in existing IT facilities.

Faced with the explosive growth of AI applications on personal computers, Intel not only provides powerful hardware, but also supports more open source models to run on personal clients through scalable software stacks. On the one hand, the 13th generation Intel ®Core processor enables large language models with up to 16 billion parameters to run on personal computers through XPU acceleration, low-bit quantization, software optimization, and support for 16GB and above memory capacity. While Intel ®®Xe graphics card and Intel sharp flash graphics card can steadily support deep learning model reasoning of FP16 precision through hardware optimization and precision control. This is of great significance for accelerating deep learning workload and improving computing efficiency. On the other hand, Intel also supports a general AI framework through an extensible software stack, which makes it easy for users to reason models in different frameworks.

Although Intel Xeon ®scalable ®processors can run many AI workloads, heterogeneous computing methods and different computing architectures are needed to support larger model scale and meet a wide range of system requirements. For these application scenarios, the Intel ®data Center GPU Max Series and Gaudi Deep Learning Accelerator provide hardware support. Among them, Intel's new Gaudi2 processor was launched in the Chinese market in July this year. As an important part of Intel's cloud-to-end product portfolio, Gaudi2 is designed for deep learning to accelerate AI training and reasoning with multi-faceted performance and energy efficiency improvements. It is worth noting that its excellent performance has also been fully verified in the recently released MLCommons ®MLPerf ®benchmark 1, which is a better solution for large-scale deployment of AI than other similar products. Thanks to the optimization of the SynapseAI ®software suite, Gaudi2 also has a unique advantage in simplifying the development and migration of models.

For a long time, Intel has been actively promoting technological transformation in China, not only accelerating the penetration of PC in China, but also promoting the development of cloud computing and edge computing and enabling the digital transformation of the industry through in-depth cooperation with local ecological partners. At a time when the wave of artificial intelligence is sweeping, Intel will also make use of its own profound accumulation and cutting-edge insight to seize new opportunities for the development of the industry together with ecological partners, so as to inject a steady stream of power into the full bloom of AI in China.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.