Intel Announces New GPU D-Series Name for Crescent Island AI Accelerators
At the OCP 2026 event, Intel announced that the official name of the Crescent Island AI accelerator is GPU D-Series and shared its technical specifications.
At the OCP 2026 event, Intel announced that the official name of the Crescent Island AI accelerator is Intel GPU D-Series, sharing details on the chip's performance and memory capacity with the public for the first time.
New Era for Crescent Island
During the OCP 2026 event, Intel announced the official name of the long-awaited Crescent Island project, stating that the chip will henceforth be known as the GPU D-Series.
Performance and Memory Data
According to the announcement made by the company, the new AI accelerator chip delivers a theoretical peak performance of 1.3 PFLOPS in FP8 computations while providing a memory bandwidth of 1.5 TB/s.
Architecture and Technical Specifications
Crescent Island holds the distinction of being the first chip built upon Intel's Xe3P graphics architecture. Its design adopts a parallel computing-focused structure instead of traditional graphics units.
Market Launch and Hardware Structure
The new GPU D-Series products will be released in an air-cooled PCIe form factor. The chips will feature a 350W TDP rating and up to 480 GB of LPDDR5X memory capacity.