Intel® Xeon® 6 Architecture - Performance and Efficiency Cores

Intel® Xeon® 6 processors offer a versatile, dual-architecture strategy featuring Performance-cores (P-cores) for compute-intensive tasks and Efficient-cores (E-cores) for high-density, scalable workloads. Built on a common platform, these processors provide up to 288 cores (E-cores) for superior performance-per-watt, ideal for cloud-native applications, or high-core-count, high-frequency (P-cores) for AI and HPC.  The new architecture on Intel® Xeon® 6 Processors is divided into P-core - Performance and E-core - Efficiency, with gains up to 2-3x on performance. 

Intel Xeon 6 P-cores (Performance-Cores)

Intel Xeon 6 processors with P-cores are optimized for high performance per core. With more cores, double the memory bandwidth, and AI acceleration in every core, Intel Xeon 6 processors provide twice the performance for the widest range of workloads, including AI and high-performance computing (HPC).4,6 Intel Xeon 6 processors with P-cores excel at a wide range of workloads, delivering better performance than any other general-purpose CPU for compute-intensive workloads like AI inference and machine learning (ML). Intel Xeon 6 processors with P-cores are great for public cloud workloads with improved performance per vCPU for floating-point operations, transactional databases, and HPC workloads. Through their leadership in AI inferencing, Intel Xeon processors continue to be the host CPU of choice on the world’s most powerful AI accelerator platforms for data preprocessing support.

• Enable AI everywhere with AI acceleration in every core. Intel® Advanced Matrix Extensions (Intel® AMX) speeds up inferencing for INT8 and BF16 and offers new support for FP16- trained models with up to 2,048 floating point operations per cycle per core for INT8 and 1,024 floating point operations per cycle per core for BF16/FP16.
• Improve memory throughput with the fastest DDR5 memory available, MRDIMM. These can deliver more than 37 percent greater memory bandwidth than RDIMMs, with an expected data transfer rate of up to 8,800 megatransfers per second (MT/s). Intel Xeon 6 processors (both P-cores and E-cores) also support DDR5 6400 high-speed memory, providing memory bandwidth gains.
• Take advantage of up to 128 cores per socket with up to 504 MB L3 cache and exceptionally low latency at large L3 access sizes. Intel® Advanced Vector Extensions 512 (Intel® AVX-512) is only supported on Xeon 6 processors with P-cores and can be used out of the box, boosting the speed of vector math common to HPC and classical AI workloads.

Intel Xeon 6 E-cores (Efficient-Cores)

Intel Xeon 6 processors with Efficient-cores (E-cores). Intel Xeon processors with E-cores are optimized for high core density and exceptional performance per watt, delivering distinct advantages for cloud-scale workloads that demand high task-parallel throughput. In comparison to 2nd Gen Intel Xeon Scalable processors, which make up the majority of systems within today’s data center footprint and are excellent candidates for performance per watt upgrades in competitive data centers, Intel Xeon 6 processors with E-cores can deliver more than 2.6x better results. This efficient performance is also ideal where power, space, and cooling are limited. Intel Xeon 6 processors with E-cores can:

• Replace four 2nd Gen Intel Xeon Scalable processor–based servers with a single server at a similar performance.
• Consolidate three racks of 2nd Gen Intel Xeon Scalable processor–based systems into a single rack.
• Accommodate AI inferencing and vector-oriented operations with Intel® Advanced Vector Extensions 2 (Intel® AVX2) and new enhancements such as Vector Neural Network Instructions (VNNI) and fast-convert for BF16 and FP16.
• Offer up to 288 cores per socket, with as much as 216 MB L3 cache, and with exceptionally low latency at large L3 access sizes.

Shared Architecture Features

Compatibility: Both types utilize the same platform and socket, allowing for infrastructure flexibility.

AI Acceleration: Both feature built-in AI acceleration (Intel® AVX and Intel® AMX/VNNI).

Security: Both include advanced security features, including Intel® TDX for confidential computing.

The efficiency of Intel Xeon 6 processors with P-cores and E-cores is highlighted by their ability to provide scalable performance per watt as server utilization increases, delivering nearly linear power-performance consumption across the load line. For performance-demanding workloads, this means power is efficiently utilized at high loads to finish jobs faster. For a scalable implementation, common to cloud or shared computing environments, this level of efficiency means that servers are only consuming the power they need when under load, reducing costs when instances are not fully utilized. The sustainability of these processors is further enhanced through system-wide power management and telemetry capabilities. These capabilities allow for increased performance per watt on a per-application basis to help with lowering overall energy consumption.

For more details, visit The Intel® Xeon® 6 Processor Family.