Intel D270 GPU AI Accelerator–Up to 480 GB LPDDR5X
Intel is pitching D270 as a memory-heavy inference card for long-context agentic workloads, not training.
The card is listed with up to 480 GB of LPDDR5X memory, 350 W air-cooled power, PCIe 5.0, 32 Xe3P cores, and 256 XMX matrix engines. Intel says the design targets prefill-heavy serving, larger contexts, concurrent sessions, and keeping more model state resident without HBM. The page frames it for enterprise AI assistants, RAG, coding assistants, document understanding, multimodal workloads, diffusion models, and high-concurrency serving. Source: HN · Frontpage AI's note.
The card is listed with up to 480 GB of LPDDR5X memory, 350 W air-cooled power, PCIe 5.0, 32 Xe3P cores, and 256 XMX matrix engines. Intel says the design targets prefill-heavy serving, larger contexts, concurrent sessions, and keeping more model state resident without HBM. The page frames it for enterprise AI assistants, RAG, coding assistants, document understanding, multimodal workloads, diffusion models, and high-concurrency serving. Source: HN · Frontpage AI's note.
score 5