With the acceleration of global digitalization and intelligence, data centers have become a key infrastructure that carries computing needs. Recently, Intel released the Xeon 6 performance core processor, which is advancing the path of change in the data center field through its own innovation. This processor brings barrel-level improvements in multi-core performance, memory bandwidth, AI acceleration, and more, especially suitable for modern AI and big data application scenarios.
Barrel upgrade
"Intel Xeon 6 performance core processors with more powerful, double the cores, faster memory, more memory, more security, and better suited for AI." Chen Baoli, vice president and general manager of Intel's Data Center and Artificial Intelligence Group in China, said at the product launch conference.
In computing, the traditional "barrel effect" means that the performance of a system is determined by the shortest block. For data centers, processor upgrades are not just about the number of cores, but also about other factors such as memory bandwidth, latency, and so on. In the Xeon 6 performance core processor, Intel has made simultaneous improvements in CPU and memory to achieve the optimal performance of the system.
The Intel Xeon 6900P series has three compute units and two IO units, and it will have up to 128 x86 cores, doubling the number of cores compared to the 64 cores of its predecessor. This means that the processor is able to handle more concurrent tasks, especially in compute-intensive applications such as AI and databases, significantly improving single-core performance and enhancing overall computing power.
In addition to the increase in the number of cores, the improvement in memory performance is also extremely important. Intel Xeon 6 performance cores support DDR5 memory with a maximum transfer rate of 6400MT/s, a significant increase in memory bandwidth compared to the previous generation's 5600MT/s. In addition, Xeon 6 introduces MRDIMM technology, which supports data transfer rates of up to 8800MT/s, providing more support for AI and scientific computing scenarios that require extremely high memory bandwidth.
The memory expansion capability has also been improved. Through CXL2.0 technology, Xeon 6 breaks through the limitations of physical memory expansion, bringing greater flexibility and scalability to data centers. This means that even as the number of CPU cores increases, the memory bandwidth and capacity can increase accordingly, ensuring that the overall performance of the system is not affected by memory bottlenecks.
The best nose experience
The Xeon 6 performance core processor not only significantly improves the number of cores and memory bandwidth, but also introduces a series of innovative technologies to further optimize AI acceleration capabilities and the overall data center experience.
AI workloads are taking on an increasingly important place in the modern data center. Intel has introduced the Advanced Matrix Extensions (AMX) instruction set in its 4th Gen Xeon processors, enabling processors to better perform AI inference tasks. In Xeon 6, the AMX instruction set has been further upgraded to provide more efficient computational support for large model inference. Especially in the inference process of generative AI models such as Llama2 and Llama3, the Xeon 6 performance core processor has a 2-3x improvement in single-core performance and performance per watt compared to the previous generation. This significant performance gain makes the Xeon 6 ideal for AI inference and training workloads.
Xeon 6 not only improves compute and memory performance, but also further improves the head-of-the-data experience. In AI accelerators such as GPUs, the CPU usually plays the role of "head", responsible for managing I/O and memory operations. With its powerful single-core performance and rich I/O expansion capabilities, the Xeon 6 is the best choice for AI acceleration systems. Working with major GPU vendors and AI ecosystem partners, Intel defined the Xeon 6 specifications to ensure it delivers the best possible headset experience.
Xeon 6 also supports secure computing, providing a strong guarantee for the security of data centers. Intel TDX2.0 technology is integrated into the processor to ensure the privacy and security of user data through hardware-level security isolation mechanisms. This technology has been widely used by many cloud vendors in China to ensure that it provides hardware-level security protection without modifying the application.
In response to the current trend of liquid cooling, Intel has cooperated with liquid cooling manufacturers to develop a variety of liquid cooling solutions. In the Xeon 6 performance core processor, cold plate liquid cooling and immersion liquid cooling technology are supported, which not only reduces the energy consumption of the data center, but also effectively improves the heat dissipation efficiency. For example, through the immersion liquid cooling solution launched in conjunction with Green Cloud Atlas, the Xeon 6 processor can maintain a stable temperature during high-performance operation, ensuring its stability and reliability in long-term operation.
Drive data center transformation
According to IDC's research, the improvement of the computing power index is positively correlated with the growth of the digital economy. The launch of Xeon 6 performance core processors will provide strong support for the improvement of computing power in data centers around the world. Its 128-core high-density computing power, coupled with significant increases in memory bandwidth and capacity, enables the Xeon 6 to efficiently handle complex AI and big data workloads. This not only accelerates the digital transformation of traditional industries, but also promotes the rapid development of emerging industries.
"Today, when cloud computing, big data, and AI have become common production tools, Intel Xeon 6 performance core processors can lay a profound foundation for us to jointly explore the transformation of data center architecture, so as to unleash the value of computing power." Yali Liang, vice president of Intel Marketing Group and general manager of Cloud & Industry Solutions and Data Center Sales in China, said.
Overall, the Xeon 6 performance core processor has achieved comprehensive improvements in computing density, memory capacity, data processing capacity, and energy efficiency optimization, which coincides with the evolution and reconstruction of data centers.
In terms of computing density, the Xeon 6's CPU cores have been increased to 128 cores, and the performance per watt has increased by 1.9 times at 40% utilization, greatly enhancing the processing power of intensive tasks such as AI, data analysis, and scientific computing. Secondly, Xeon 6 introduces MRDIMM and CXL2.0 technologies, which significantly improve memory bandwidth and capacity, and solve the memory performance matching problem caused by the increase in the number of CPU cores.
In addition, in terms of data processing and security, Xeon 6 enhances the accelerator technology by integrating TDX and SGX security mechanisms to ensure the high speed and security of data processing, which is especially suitable for confidential computing and trusted service application scenarios. Finally, Xeon 6 provides strong support for the construction of low-carbon data centers by optimizing CPU energy consumption and combining advanced cooling technology to effectively cope with high computing power demand and reduce energy consumption.
Intel Xeon 6 performance core processors have demonstrated leading technological advantages in terms of CPU cores, memory bandwidth, and AI acceleration and security. Xeon 6 will become an important engine to drive data center transformation, providing optimal computing support for AI and big data applications, and providing innovative solutions for building low-carbon and green data centers.