Intel began customer sampling of D-Series inference GPUs; the upcoming D270 specifies 1.3 petaflops dense FP8 and 1.5 TB/s memory bandwidth.

Intel’s previously identified “Crescent Island” data-center AI inference cards now have the official Intel GPU D-Series branding. ServeTheHome reports that Intel published performance specifications for the upcoming D270, while financialjuice separately reports that customer sampling has begun.

The D270 specification includes 1.3 petaflops of dense eight-bit floating-point (FP8) performance. It also specifies 1.5 terabytes per second of memory bandwidth, a separate measurement from arithmetic throughput.

Customer sampling establishes an evaluation-stage milestone, not general availability or broad production deployment. The supplied sources do not disclose pricing, customer identities, memory capacity, power consumption, shipment volumes, or measured application-level inference results.