Intel Xeon 6 CPUs launched with AI-first features: All the details

Intel has launched three new processors in the Intel Xeon 6 series with Performance-cores (P-cores), optimised for GPU-accelerated artificial intelligence workloads. These chips are now available globally and are designed to deliver enhanced CPU-GPU synergy, as demonstrated by their inclusion in NVIDIA’s DGX B300 AI systems.

Features
The headline feature of the new Xeon 6 processors is Intel’s Priority Core Turbo (PCT) technology. It enables certain CPU cores to dynamically scale to higher turbo frequencies, while other cores maintain base performance. This prioritization enhances the CPU’s ability to feed data efficiently to GPUs handling large-scale AI tasks. The chips also include Intel Speed Select Technology – Turbo Frequency (SST-TF) for customizable frequency scaling, further optimizing power and thermal efficiency in AI compute environments.

Intel highlighted several additional specifications, including:
• Core Count: Up to 128 Performance-cores per CPU, enabling efficient parallel processing.
• Memory Performance: Up to 30% faster memory speeds compared to competing solutions, using MRDIMMs and Compute Express Link (CXL).
• I/O Capacity: Up to 20% more PCIe lanes than previous-generation Xeon processors for improved data throughput.
• AI-focused Extensions: Support for Intel Advanced Matrix Extensions with FP16 arithmetic, enabling preprocessing of AI data directly on the CPU.

One of the new chips, the Intel Xeon 6776P, is currently shipping inside the NVIDIA DGX B300 system. This model serves as the host CPU, responsible for orchestrating AI operations and managing data transfer between components in the system.