Intel announces its GPU D-Series, enterprise-focused AI accelerator cards built for inference workloads, as part of its “Crescent Island” data center effort. The announcement is made in connection with industry events including the Open Compute Project (OCP) summit, where Intel presents details of the accelerators aimed at running AI models rather than training.
Across coverage, the products are described as inference-optimized and branded under the “GPU D-Series” name. ServeTheHome adds additional technical specifics for a model referred to as the D270, including performance figures for dense FP8 and stated memory bandwidth, positioning the cards for high-throughput serving scenarios in data centers. Other outlets focus more on the naming and overall announcement rather than granular specifications. No competing performance comparisons or pricing are reported in the provided materials.
