Intel GPU D270(Crescent Island)Intel GPU D270 (Crescent Island)
未公布单价No public price未公布单价No public price · 官方页未给公开单价;这类产品通常按整机、云实例或询价成交,不拿经销商价冒充官方价。No public unit price on the official page; these usually sell as systems, cloud instances or on request, so dealer quotes are not presented as official.
回答「不想上机柜级液冷、又要在标准 PCIe 服务器里跑大显存推理,Intel 现在有什么卡」。D270 是 Intel 当前的数据中心推理 GPU:Xe3P、32 个 Xe core / 256 个 XMX、160GB LPDDR5X、1.5TB/s、350W、PCIe 5.0 风冷,原生 FP4 到 FP64;单卡可跑 300B 参数,8 卡可跑 1T+。它的差异点不是峰值算力,而是「大显存 + 低功耗 + 标准形态」。
不主张:480GB 是合作方配置,不是 Intel 品牌卡默认容量;1.3 PFLOPS 是 FP8,不是 FP4;单卡 300B / 8 卡 1T+ 是官方容量口径,不等于这些模型能以理想吞吐运行;官方给的性能对比基于 pre-production 芯片,属于厂商测试,不是独立实测;没有公开价格与正式上市日期。Answers what Intel offers when a standard PCIe server must run large-memory inference without rack-level liquid cooling. D270 is Intel's current data-center inference GPU: Xe3P, 32 Xe cores / 256 XMX engines, 160 GB LPDDR5X at 1.5 TB/s, 350 W, PCIe 5.0 air-cooled, native FP4 through FP64; one card holds up to 300B parameters and eight cards more than 1T. Its differentiator is large memory plus low power in a standard form factor, not peak compute.
It does not claim: 480 GB is a partner configuration, not the default Intel-branded card; 1.3 PFLOPS is FP8, not FP4; the 300B / 1T+ figures are capacity statements, not a promise of ideal throughput; Intel's comparisons use pre-production silicon and are vendor tests, not independent measurements; no public price or general-availability date.
关键参数Key specs
- 架构 / architecture
- Intel Xe3P
- Xe cores / XMX
- 32 Xe cores / 256 XMX engines
- 显存 / memory
- 160 GB LPDDR5X(合作方配置可至 480 GB)
- 带宽 / bandwidth
- 1.5 TB/s
- FP8 算力 / FP8 compute
- 1.3 PFLOPS
- 精度 / precision
- 原生 FP4–FP64,FP64 全速
- 功耗 / power
- 350 W
- 形态 / form factor
- PCIe 5.0 插卡,风冷
- 模型容量 / model capacity
- 单卡 300B;8 卡 1T+
- 状态 / status
- 2026-11-16 官方 HPC 文章;OCP 2026 公布 D 系列
官方原文怎么说What the source says
Silicon & Precision: Xe 3P architecture, 32 Xe cores, 256 XMX engines; native FP4 to FP64 with full-rate double precision.
Memory & Platform: 160 GB LPDDR5X (1.5 TB/s bandwidth, scalable up to 480 GB); 350W PCIe 5.0 air-cooled form factor.
The GPU combines up to 480 GB of LPDDR5X memory, 1.5TB/s of memory bandwidth and 1.3 petaflops of FP8 compute to help improve the economics of sustained, memory-intensive AI inference.
supports models with up to 300 billion parameters on a single card and more than one trillion parameters in an eight-GPU system.
资料与许可Source and licence
- 资料时间Source checked
- 2026-11-20
- 许可 / 类型Licence / kind
- official-spec-page
- 许可说明Licence evidence
- https://www.intel.cn/content/www/us/en/newsroom/news/data-center/why-intel-gpu-d270-belongs-in-the-hpc-conversation.html
- 类目Category
- 加速卡Accelerators
- 状态Status
- 已发布 / 未出货Announced
- 价格口径Price basis
- 未公布单价No public price · not-published