Shulou(Shulou.com)12/24 Report--

Today, at the 2023 Intel new product launch cum AI technology innovation party with the theme of "AI is everywhere", Intel officially launched the fifth generation Intel ®Xeon ®Extensible processor (code name Emerald Rapids). During this period, Intel also shared with its ecological partners the successful practice and application value of the new product in Jingdongyun, Baidu Smart Cloud, Aliyun and Volcano engine.

The fifth generation Intel ®Xeon ®scalable processors excel in improving ① per watt performance of critical workloads such as artificial intelligence, scientific computing, networking, storage, database, security, and reducing total cost of ownership (TCO). ② is the second release of the Intel Xeon scalable processor family within a year. Compared with the previous generation, the fifth generation Intel ®Xeon ®scalable processors can provide higher computing power and faster memory in the same thermal design power consumption (TDP) range. At the same time, the ② processor is compatible with the software and platforms of the previous generation, enabling customers to upgrade and significantly increase the life of their infrastructure while reducing costs and carbon emissions.

Sandra Rivera, executive vice president and general manager of Intel's data center and artificial intelligence business, said, "the fifth generation Intel Xeon scalable processors created for AI acceleration are designed to provide higher performance for customers deploying artificial intelligence in cloud, network, and edge environments. Thanks to our long-term cooperation with the ecosystem of customers, partners and developers, we have launched the fifth-generation Intel Xeon scalable processor on a proven basis. this will enable customers and partners to rapidly apply and scale at a lower total cost of ownership. "

At the press conference, Chen Baoli, vice president and general manager of Intel data Center and artificial Intelligence Group, also pointed out in the speech, "AI is not only changing our way of life, but also bringing changes to the development and innovation of each industry. As a leader in technological innovation, Intel is committed to enabling more customers to design, deliver and deploy innovative solutions for cloud computing, networking, edge and large-scale access businesses through a comprehensive product portfolio, optimized software, diverse tools and a broad ecosystem. "

The fifth-generation Intel ®strong ®scalable processor has AI acceleration in each core, which can handle demanding end-to-end AI workloads without adding independent accelerators, including a 42% improvement in reasoning performance and a latency of less than 100ms for large language models with up to 20 billion parameters. three

Compared to the previous generation, the fifth-generation Intel ®Xeon ®scalable processors can increase overall performance by up to 21% ④ for general-purpose computing and up to 36% performance per watt in a range of customer workloads ⑤. For customers who follow a typical five-year update cycle and upgrade from older processors, the total cost of ownership is reduced by up to 77 ⑥.

Compared with previous generations, the fifth generation Intel ®Xeon ®scalable processors will provide customers with more innovation to improve performance and energy efficiency. The processor has as many as 64 cores and the three-tier cache is nearly three times that of the previous generation. At the same time, the processor has eight DDR5 channels, supports speeds of up to 5600 MT/s, and uses Intel ®®UPI 2.0 to increase cross-slot bandwidth and provide up to 20 GT/s of transmission. In addition, leading cloud service providers (CSP) will be able to use CXL Type 3 memory devices to expand memory capacity with instances based on fifth-generation Intel ®Xeon ®scalable processors.

The fifth-generation Intel ®strong ®scalable processor is pin-compatible with the previous generation of products. Large original equipment manufacturers (OEM), including Cisco, Dell, HPE, Tide, Lenovo, and ultra-micro computers, will offer multiple single-core and dual-core processors for the market to choose from in the first quarter of 2024, while mainstream CSP will gradually launch examples based on this new product next year.

Intel ®trusted Domain extension (Intel ®TDX) provides isolation and confidentiality at the virtual machine (VM) level, thereby enhancing privacy and data management. Intel ®TDX has been integrated into the fourth-generation Intel ®Xeon ®extensible processor released earlier this year, and specific CSP has been able to apply this feature, which can be enabled by all OEM and CSP solution providers with the introduction of new products. In secret virtual machines based on Intel ®TDX, client operating systems and virtual machine applications are isolated from being accessed by cloud hosts, hypervisors, and other virtual machines on the platform.

This new Intel Xeon scalable processor provides significant performance improvements and overall benefits to customers and partners, and enables performance optimization for critical workloads.

The fourth and fifth generation Intel ®Xeon ®scalable processors were launched in January this year and today, respectively, which indicates that Intel is continuing to promote the development of its own product roadmap and fulfill its commitment to customers. Looking ahead, the next generation of Intel Xeon scalable processors to be launched next year will bring significant improvements in energy efficiency and performance. Sierra Forest, an energy efficient core (E-core) processor with up to 288 cores, will be launched in the first half of 2024, followed by the performance core (P-core) processor Granite Rapids.

Note:

The average performance improvement of ①, as measured by the geometric averages of SPEC CPU rates, STREAM Triad and LINPACK, compared to fourth-generation Intel Xeon processors. See G1 on intel.com/processorclaims: fifth-generation Intel Xeon processors. The results may be different.

Compared with the fourth-generation Intel Xeon processors, ② is measured by the performance per watt of a range of artificial intelligence, database, network, and scientific computing workloads. See A2, A19-A25, D1, D2, D5, H1, N16: fifth-generation Intel Xeon processors on intel.com/processorclaims. The results may be different.

③ is based on the Intel internal model as of December 2023. See A1, A2, A16 on intel.com/processorclaims: fifth-generation Intel Xeon processors. The results may be different.

The average performance improvement of ④, as measured by the geometric averages of SPEC CPU rates, STREAM Triad and LINPACK, compared to fourth-generation Intel Xeon processors. See G1 on intel.com/processorclaims: fifth-generation Intel Xeon processors. The results may be different.

For ⑤, see [T13]: fifth-generation Intel Xeon Extensible processors on intel.com/processorclaims. The results may be different.

For ⑥, see [T7]: fifth-generation Intel Xeon Extensible processors on intel.com/processorclaims. The results may be different.

Figure 2:

AI reasoning: see [A25]: fifth Generation Intel Xeon Extensible processors on intel.com/processorclaims. The results may be different.

AI reasoning: see [A26]: fifth Generation Intel Xeon Extensible processors on intel.com/processorclaims. The results may be different.

Scientific computing: see [H7]: fifth-generation Intel Xeon Extensible processors on intel.com/processorclaims. The results may be different.

Network: see [N1]: fifth Generation Intel Xeon Extensible processors on intel.com/processorclaims. The results may be different.

Web and Cloud Native: see [W3]: fifth Generation Intel Xeon Extensible processors on intel.com/processorclaims. The results may be different.

Infrastructure and storage: see [N16]: fifth Generation Intel Xeon Extensible processors on intel.com/processorclaims. The results may be different.

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.