Intel Xeon 6 processors: Drive data center integration and upgrades with superior performance and power efficiency

Following the launch of the Intel Xeon 6900 performance core processor in September last year, Intel further expanded the Xeon 6 product family, and recently released a number of new products including Xeon 6700 performance core processor and Xeon 6500 performance core processor, with a richer product portfolio, superior performance and excellent energy efficiency. Address the demands of a wide range of workloads across the data center, network and edge.

Rich Xeon 6 product portfolio

Intel Xeon 6 performance core processors are designed for compute-intensive workloads and offer highly flexible configuration options. From a minimum of 16 cores to a maximum of 128 cores, customers can choose the best configuration according to the actual application. For traditional large-scale databases and other scenarios that have high requirements for connection speed and multi-core parallel processing capabilities, the new Xeon 6700/6500 performance core processor not only brings significant improvements in the number of cores, I/O bandwidth and memory capacity, but also adds support for four-way and eight-way server operation, fully releasing multi-core performance through high-speed UPI interconnection. To better meet customer needs for high-performance database applications.

Like the Xeon 6900 performance core processor with up to 128 cores, the Xeon 6700/6500 performance core processor also supports up to 6400MT/s DDR5 memory, 8800MT/s MRDIMM memory, and 64 CXL 2.0 channels. CXL extended memory allows users to expand server memory space at a lower cost. With CXL 2.0 technology, Intel also offers a "Flat Memory Mode" feature that automatically switches between hot data and warm data stored on the CXL memory, achieving a balance between performance and capacity.

For enterprises that require large-scale storage solutions (SSDS), as well as platforms that rely on content delivery networks (CDN) to deliver a smooth video experience, I/O connectivity is important. In addition, AI servers need to connect multiple Gpus, which also puts higher requirements on CPU and platform I/O scalability. The Xeon 6700/6500 performance core processor supports a single slot platform with up to 136 PCIe lanes, providing strong scalability and meeting the need for high connectivity.

The nose node of the data center

In modern AI systems, GPU and CPU both play a very important role. Head-node cpus need not only excellent I/O speed, powerful single-threaded performance, and high memory bandwidth and capacity, but also the reliability, availability and maintainability (RAS) requirements necessary for large systems to operate 24/7. In addition, flexible support for a variety of forms and configurations is also important to optimize data center space utilization, improve heat dissipation efficiency, and adapt to diverse workloads.

As a node-node CPU for AI systems, Xeon 6 can be paired with Gpus to leverage the full potential of Gpus by delivering faster I/O speeds and better single-core performance. In a typical dual-CPU head-node configuration, the Xeon 6700 performance core processor platform offers up to 172 CPU cores (344 threads) and is equipped with 16 DDR5 high-speed memory channels (32 memory slots) and 176 PCIe 5.0 links (128 of which can be configured as CXL 2.0 links).

Powerful AI reasoning performance

In terms of AI reasoning, Xeon 6 processors with the built-in AI acceleration engine AMX (Intel Advanced Matrix Extension) excellent capabilities, can allow the system with fewer cores, get up to 1.5 times AI reasoning performance improvement 1. This advantage makes it great in traditional machine learning, small generative AI models, and GPU-accelerated workloads. For traditional AI applications, such as recommendation engines, voice assistants, image recognition, etc., Xeon 6 can provide powerful computing support. Even in the face of small generative AI models with parameters less than 20B, Xeon 6 still shows excellent performance.

In addition, for AI reasoning of small language models, not only requires high-intensity computation when generating the first Token, but also has extremely high memory bandwidth requirements for subsequent reasoning processes. Xeon 6 processors are the first in the industry to support MRDIMM technology, providing ultra-fast memory up to 8800MT/s. With the support of MRDIMM technology, the bandwidth is doubled, enabling AI workloads such as small LLM, traditional deep learning and recommendation systems to run smoothly on Xeon platforms, greatly unlocking the potential of Xeon 6 performance core processors.

With excellent performance and energy efficiency, help data center integration and upgrading

Driven by AI, the demand for computing power is growing at an unprecedented rate, and at the same time, the demand for CPU computing power in data centers is also growing at the same time, and there is an urgent need to meet new challenges by integrating and upgrading. Unlike typical update cycles, which typically focus on upgrading hardware, data center consolidation is deeper and more focused on optimizing existing resources to reduce costs, improve scalability, and achieve sustainability. As a result, enterprises are not only focusing on performance improvement when it comes to infrastructure renewal, but also paying great attention to space, energy consumption and cost savings.

With superior performance and power efficiency, Xeon 6 processors provide the ideal solution for data center consolidation, not only delivering higher computing power with fewer servers, but also significantly reducing energy consumption and total cost of ownership (TCO). Taking the second generation Xeon, the most widely used Xeon processor, as an example, with its excellent performance efficiency per watt, the Xeon 6 processor can replace five old servers on average in data services workloads, and even one new server can replace ten old servers in image recognition and network services workloads. This helps reduce TCO3 by up to 68%.

In the face of increasingly diverse and quantified computing power needs, Intel Xeon 6 series processors have been widely used in the data center ecosystem with superior performance and excellent energy efficiency. Currently, more than 500 Xeon 6 processor-based product designs have been launched or are under development. Intel not only provides enterprises with powerful and easy-to-use AI systems through open solutions based on x86 architecture, across hardware and software, but also works with a wide range of ecological partners to build an AI ecosystem to accelerate workloads from the data center to the cloud, edge and PC with better performance per watt and lower TCO.