Intel D270 GPU AI Accelerator–Up to 480 GB LPDDR5X
Intel's air-cooled D270 data center GPU pairs 480 GB LPDDR5X with inference-focused Xe3P compute.
Intel's GPU D270 is an inference-oriented data center accelerator with up to 480 GB of LPDDR5X and a 350 W air-cooled design. It provides 32 Xe3P cores and 256 XMX matrix engines on a standard PCIe 5.0 card and does not use high-bandwidth memory. Intel positions it for agentic workloads with long contexts and many concurrent sessions, served through vLLM and PyTorch on existing racks. The page is a product overview and signup form rather than an independent benchmark report.
- Up to 480 GB of LPDDR5X per card, with no HBM.
- 32 Xe3P cores and 256 XMX engines in a 350 W card.
- Standard air-cooled PCIe 5.0 form factor for existing servers.
- Positioned for long-context agentic inference on vLLM and PyTorch.
- Vendors
- Intel
- Products
- Intel GPU D270Intel GPU D-series
- Organizations
- Intel
From the original article
8 min read at intel.comAgents read far more than they write: long contexts, multistep workflows, compute dominated by prefill. As demands grow, challenges expand beyond bandwidth. Intel D-series GPUs are built for this new era of agentic AI, efficiently handling larger contexts, more concurrent sessions, and more tokens within practical power and cost limits.
By hurricanepootis · intel.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.