Intel details 350-watt D-series GPU for enterprise AI inference

Intel disclosed new specifications for its D-series GPU at the Open Compute Project Global Summit, describing a 350-watt, air-cooled PCIe card for running AI models, SiliconANGLE reported. Intel says the Xe3P design delivers 1.3 petaflops of FP8 compute and 1.5 terabytes per second of memory bandwidth. Intel's branded card carries 160 gigabytes of LPDDR5X memory; the design supports partner cards with up to 480 gigabytes. Intel says a single card can support models with up to 300 billion parameters. The memory choice trades peak bandwidth for capacity and cost, and the product still needs to ship and perform as promised.