At the Open Compute Project (OCP) Global Summit, Intel introduced new technical details on the Intel GPU D-series, formerly code-named Crescent Island, alongside new Xeon-based rack designs and open software innovations. The announcements reinforce Intel’s strategy of combining general-purpose compute, purpose-built acceleration, and open standards to address the systems challenges of agentic AI.
Main points in press release:
- Intel unveiled new Xeon rack designs and the GPU D-series at OCP to support agentic AI workloads.
- The company positions open, heterogeneous infrastructure combining CPUs, GPUs, and interoperable software as the solution.
- Xeon 6+ configurations deliver extreme core density; GPU D-series targets efficient memory-intensive inference at scale.
Agentic AI transforms inference into multi-step workflows involving retrieval, reasoning, and tool use, placing pressure on power, memory, and cost. While open models expand enterprise choice, scaling remains a systems problem. Intel positions its stack as a solution: Xeon CPUs provide the high-density foundation, while the Xe3P-based GPU D-series targets efficient, memory-intensive inference at scale.
New Xeon 6+ configurations scale beyond 46,000 cores per rack at 133.45 kW, delivering the CPU compute density needed to orchestrate workloads and keep accelerators fed. The GPU D-series, built on Xe3P architecture, offers up to 160GB LPDDR5x memory with application portability across Intel GPU platforms, reducing the need for specialized rack or cooling infrastructure. Nice one!
[Read More]