At this year’s Hot Chips event, Intel has taken the stage to offer an in-depth preview of its forthcoming Intel Xeon product lineup, ushering in a new era for the renowned processor series with a groundbreaking platform architecture.
The platform marks a significant advancement for Intel Xeon as it introduces two distinct architectures, namely the Efficient-core (E-core) and Performance-core (P-core) architectures, with the codenames Sierra Forest and Granite Rapids, respectively. These groundbreaking products are poised to revolutionize the landscape of critical workloads, including artificial intelligence.
The introduction of Intel Xeon processors with E-cores (Sierra Forest) signifies a pivotal shift towards power-efficient computing while maintaining impressive performance levels. Designed to deliver density-optimized compute capabilities, the E-core architecture brings forth industry-leading power-performance density, setting the stage for outstanding performance in cloud-native and hyperscale workloads. Key features of the E-core architecture include:
Achieving 2.5 times better rack density and a remarkable 2.4 times higher performance per watt.
Support for 1-socket (1S) and 2-socket (2S) servers, featuring up to an impressive 144 cores per CPU with a remarkably low thermal design power (TDP) starting at just 200 watts.
Integration of a modern instruction set, fortified security measures, virtualization capabilities, and advanced vector extensions (AVX) with AI extensions.
Inclusion of foundational memory reliability, availability, and serviceability (RAS) features, ensuring machine check and data cache error correction code (ECC) across all Xeon CPUs.
In a bid to provide an unparalleled total cost of ownership (TCO) for high-core performance-sensitive workloads and general-purpose compute tasks, Intel Xeon processors with P-cores (Granite Rapids) have been meticulously designed. Granite Rapids, building on the AI performance capabilities of the Xeon series, integrates accelerators to further boost targeted workloads, enhancing both performance and efficiency. Key features of the P-core architecture include:
Delivering 2-3 times better performance for mixed AI workloads, amplifying its prowess in AI-centric tasks.
Enhanced Intel Advanced Matrix Extensions (AMX) with new support for FP16 instructions, unlocking new possibilities in AI and data-intensive computations.
Amplified memory bandwidth, core count, and cache tailored for compute-intensive workloads.
Exceptional socket scalability, ranging from single-socket to eight-socket configurations, ensuring versatility and adaptability.
