Intel GPU D-Series Enters AI Inference With 480GB LPDDR5X

When people talk about AI accelerators, they usually lead with compute numbers. For inference, the workload that runs trained models to answer questions or generate output, memory often matters more.