Intel’s Crescent Island Xe3P GPU Targets AI Inference with 160GB LPDDR5x Memory
Intel has revealed its Crescent Island data center GPU featuring the new Xe3P architecture and up to 160GB LPDDR5x memory. The processor is specifically designed for AI inference workloads with improved power efficiency. Samples are expected to ship in late 2026 ahead of full launch in 2027.
Intel Announces Xe3P Architecture for Data Center AI Workloads
Intel has unveiled significant details about its upcoming data center GPU codenamed Crescent Island, featuring the company’s new Xe3P architecture, according to reports from the OCP Global Summit. The announcement represents a strategic shift in Intel’s approach to data center acceleration, with the company positioning the technology specifically for AI inference workloads rather than full-scale training operations.