Intel focuses on agentic AI with Xeon 6+ and Ethernet enhancements

Intel has recently introduced a new range of data center hardware designed to support the growing prevalence of agentic AI. This update encompasses enhancements across CPUs, networking solutions, and upcoming AI accelerators. The company emphasizes the significance of CPUs as the central layer for orchestrating complex AI systems in the evolving technological landscape.

The announcement was made at Computex in Taipei and features the introduction of new Intel Xeon 6+ processors, Intel Ethernet E835 networking components, and detailed insights into Intel’s forthcoming Crescent Island GPU platform. This launch underscores the shift in focus among chipmakers towards optimizing infrastructure for power efficiency, data transfer, and scalable AI inference, rather than solely concentrating on raw accelerator performance.

Xeon 6+ Emphasizes AI Orchestration

The latest Xeon 6+ processors expand Intel’s Xeon 6 family by incorporating up to 288 Efficient-cores and are constructed using Intel’s 18A process technology, a first for a data center CPU. These chips are tailored for cloud-native and AI-intensive workloads where factors like concurrency, latency, and rack density are emerging as critical considerations. Intel claims that these processors can deliver up to 2.5 times higher performance compared to the previous generation, while also enhancing performance-per-watt metrics when pitted against rival products.

According to Kevork Kechichan, Executive Vice President and General Manager of Intel Data Center Group, “AI doesn’t scale as a collection of parts—it scales as a coordinated system.” He further emphasizes the importance of orchestration, concurrency, and data movement in the context of AI evolution. This underscores the pivotal role of the CPU as the control plane for modern AI infrastructure. Through Xeon 6+ and Ethernet E835, Intel aims to integrate compute and networking closely to alleviate bottlenecks and facilitate efficient, secure scaling of real-world agentic workflows.

The processors also boast support for 12-channel DDR5 memory, PCIe Gen 5, CXL connectivity, and integrated security features such as Intel SGX and TDX. Intel reports that these chips are currently undergoing testing in telecom infrastructure and are being integrated into systems by prominent vendors like ASUS, Dell Technologies, Ericsson, HPE, Lenovo, and Supermicro.

Ethernet E835 Addresses AI Networking Challenges

Concurrently with the CPU launch, Intel has expanded its Ethernet lineup with the introduction of E835 controllers and network adapters capable of supporting data rates up to 200GbE. The company underscores the growing importance of networking efficiency as AI and distributed cloud workloads expand. The E835 products support RDMA technologies like RoCEv2 and iWARP, while also focusing on reducing power consumption in dense server deployments.

Intel asserts that its E835-CQDA2 adapter offers superior performance-per-watt compared to similar products from competitors like NVIDIA and Broadcom. These adapters also integrate security functionalities such as Hardware Root of Trust and signed SPDM support.

Crescent Island GPU Progresses

Intel has also disclosed additional details about its upcoming AI accelerator, codenamed Crescent Island, which is based on the Xe 3P architecture. This GPU is tailored for handling extensive inference workloads and agentic AI systems, featuring up to 480GB of LPDDR5x memory and a 350W air-cooled PCIe design. Intel highlights that the platform supports a wide range of datatypes from FP4 to FP64 and is intended to seamlessly integrate with the company’s open software stack for heterogeneous computing environments.

Simultaneously, Intel has introduced a new 12-core Xeon 6300 processor aimed at SMB entry servers, providing smaller systems with access to higher core counts without necessitating platform redesigns.