Intel GPU D-series (Crescent Island): 160GB PCIe card, sampling, no price

On October 12, 2026, at the Open Compute Project (OCP) Global Summit, Intel shared new technical details on the Intel GPU D-series, formerly code-named Crescent Island. It is an inference GPU built on Intel's Xe3P architecture. Intel's headline figures are up to 480 GB of LPDDR5X memory, 1.5 TB/s of memory bandwidth and 1.3 petaflops of FP8 compute, in a 350-watt power envelope. A footnote narrows the memory figure: the Intel-branded PCIe card has 160GB of LPDDR5x. Partners can build ODM-branded cards with options up to 480GB. The release does not say whether the 1.5 TB/s figure applies to the 160GB Intel-branded card. The card uses a standard PCIe form factor with air cooling. Intel says it supports models of up to 300 billion parameters on one card, and more than one trillion parameters in an eight-GPU system. The release does not state the precision or the memory configuration behind those model-size figures. Intel frames the card for existing data center environments. Customer sampling is underway. The release gives no price and no general availability date. It does not name specific inference runtimes. It refers in general terms to upstream integration with industry-standard frameworks and serving engines, and to a common Xe software foundation shared with other Intel GPUs.