Intel names its Crescent Island inference accelerators GPU D-Series and details specifications
Intel introduced the GPU D-Series as the product family for its Crescent Island data-center inference accelerators. The cards are rated for up to 1.3 petaflops of FP8 compute and 1.5 TB/s of memory bandwidth, with up to 480 GB of LPDDR5X; the D270 is the first part number identified. The 480 GB capacity applies to partner-built cards, while Intel’s own card has 160 GB, and the design uses a 350W air-cooled PCIe form factor. Intel says customer sampling is underway.
