Intel Xeon 6 processors power NVIDIA DGX Rubin NVL8 systems, marking a key collaboration for AI inference workloads. NVIDIA announced at GTC 2026 that the Xeon 6 serves as the host CPU, providing architectural continuity and scalability for GPU-accelerated AI systems. This shift highlights inference's growing importance over large-scale training, driven by agentic AI and reasoning systems. Jeff McVeigh of Intel emphasized the host CPU's critical role in orchestration, memory access, model security, and throughput across GPU systems, stating Xeon 6 delivers leadership performance and compatibility with the extensive x86 software ecosystem customers rely on for scaling inference workloads.

The selection addresses modern AI infrastructure needs where inference performance is defined by both GPU throughput and CPU-led system efficiency. Xeon 6 offers fast memory speeds, balanced workload performance, lower long-term TCO, and mature enterprise software support. Its robust PCIe and I/O capabilities further strengthen its position as a high-bandwidth, low-latency platform. Key features include up to 8 TB system memory, 3X higher memory bandwidth, industry-leading PCIe 5.0 lanes, and hardware-rooted isolation via Intel Trust Domain Extensions (TDX) for confidential computing across CPU-GPU data paths.

This partnership reinforces Xeon's role as a cornerstone of modern AI infrastructure, enabling scalable deployments across data centers, cloud, and edge. The DGX Rubin NVL8 builds on the Xeon 6776P foundation used in current Blackwell platforms, ensuring performance continuity. Efficient Priority Core Turbo features help maximize GPU utilization, while strong single-thread performance handles complex orchestration and data movement, crucial as inference workloads grow more demanding.