Intel Details Data Center GPU D-Series: D270 Pairs Large Memory With 350W Air Cooling, Now Sampling to Customers
SAN JOSE — Intel shared new details of its data center GPU D-series on October 12 at the Open Compute Project Global Summit in San Jose. The line is the inference accelerator announced last October under the codename Crescent Island; it now has the D-series name, fuller specifications and confirmation that customer sampling is underway.
The first product, the D270, uses Intel's Xe3P architecture and a PCIe 5.0 connection, with a 350-watt power envelope and passive air cooling supplied by the server — avoiding the need for dedicated liquid cooling. Intel's own PCIe card carries 160GB of LPDDR5X memory; per a footnote in Intel's announcement, manufacturing partners can build cards with as much as 480GB. Memory bandwidth is 1.5 TB/s. Intel says the family can support models with up to 300 billion parameters on one card, or more than a trillion across eight GPUs. Listed uses include enterprise assistants, document processing, coding assistants, image-generation models and SI agents.
On software, the D-series supports PyTorch and vLLM and shares a software foundation with Intel's other Xe GPUs. Intel's announcement gives no pricing or general-availability date.
Intel ranks No. 4 in GSM's Global Chip Company Rankings on 2025 revenue of $47.9 billion, down 3.9% — the only decline among the top 10. Its third-quarter results are due after the market close on October 29.
Published by GSM · USA (original Chinese edition)
Photo: 图：英特尔位于加州 Santa Clara 的总部大楼。照片：Coolcaesar，CC BY-SA 4.0，经 Wikimedia Commons。Photo: Intel's headquarters in Santa Clara, California. Photo: Coolcaesar, CC BY-SA 4.0, via Wikimedia Commons.
