Intel Names Data Center Inference GPU Family 'D-series', Debuts D270 at OCP Global Summit

At the 2026 OCP Global Summit, Intel officially announced that its data center inference GPU "Crescent Island," based on the Xe3P architecture, will be branded as the Intel GPU D-series, with the Intel GPU D270 as the first product.

The Intel GPU D270 features 32 Xe cores and 256 XMX engines, delivering 1.3 petaflops of FP8 compute. It supports up to 480GB of LPDDR5X memory with 1.5TB/s bandwidth, has a 350W total board power, and comes in a standard PCIe AIC form factor with PCIe Gen5 support and air-cooled design capabilities.

The D-series GPUs are designed for LLM inference, long-context workloads, and concurrent inference serving. A single card supports models up to 300B parameters, while an 8-card system can support 1T-parameter models. Target use cases include multi-tenant AI services, robotics, and other latency-sensitive workloads.

Intel also showcased new rack designs based on Xeon processors with partners. Some configurations using Xeon 6+ can scale to more than 46,000 CPU cores with power draw up to 133.45kW.