当前位置:首页 > 报告详情

利用光互连扩展人工智能基础设施.pdf

上传人: 明**** 编号:1011763 2025-12-21 17页 1.18MB

1、Scaling AI Infrastructure with Optical InterconnectsJosef BergerAVP Marketing,Cloud OpticsOctober 14,2025|OCP 2025 2025 Marvell.All rights reserved.2OCP 2025 October 14,2025Forward-looking statementsExcept for statements of historical fact,this presentation contains forward-looking statements(within

2、 the meaning of the federal securities laws)including statements related to future revenue,future earnings,and the success of our product releases that involve risks and uncertainties.Words such as“anticipates,”“expects,”“intends,”“plans,”“projects,”“believes,”“seeks,”“estimates,”“can,”“may,”“will,”

3、“would”and similar expressions identify such forward-looking statements.These statements are not guarantees of results and should not be considered as an indication of future activity or future performance.Actual events or results may differ materially from those described in this presentation due t

4、o a number of risks and uncertainties.Forward-looking statements are only predictions and are subject to risks,uncertainties and assumptions that are difficult to predict,including those described in the“Risk Factors”section of our Annual Reports on Form 10-K,Quarterly Reports on Form 10-Q and other

5、 documents filed by us from time to time with the SEC.Forward-looking statements speak only as of the date they are made.You are cautioned not to put undue reliance on forward-looking statements,and no person assumes any obligation to update or revise any such forward-looking statements,whether as a

6、 result of new information,future events or otherwise.2025 Marvell.All rights reserved.3OCP 2025 October 14,2025 1M XPUs across multiple sites 10 Exabit/s bandwidth between sites 1 Gigawatt campus Exascale AI factories of tomorrow 2025 Marvell.All rights reserved.4OCP 2025 October 14,2025CopperOptic

word格式文档无特别注明外均可编辑修改,预览文件经过压缩,下载原文更清晰!
三个皮匠报告文库所有资源均是客户上传分享,仅供网友学习交流,未经上传用户书面授权,请勿作商用。
根据《Scaling AI Infrastructure with Optical Interconnects》的内容,以下是全文关键点的概括: 1. **多站点AI工厂规模**:计划在多个地点部署1M个XPU,实现10 Exabit/s的站点间带宽,以及1 Gigawatt的校园级数据中心。 2. **数据中心互联(DCI)**:采用铜缆和光学互联技术,实现从1米到超过1000公里的可扩展连接。 3. **互联层级**:包括铜缆、短距离光纤和长距离光纤三个层级,以满足不同距离和功率需求。 4. **光学互联技术**:采用CPO(共封装光模块)技术,实现超过1000公里的可扩展性。 5. **PAM4技术**:采用PAM4可插拔模块,每两年带宽翻倍,目前已有400G/通道PAM4电光转换技术。 6. **Coherent-lite技术**:结合相干和IMDD光学技术,实现更高带宽和更远距离的连接。 7. **多站点集群**:通过ZR/ZR+和Coherent-lite光学模块,实现多站点集群的扩展。 8. **关键数据**:32 x 224G的光学通道,每台引擎6.4T,每根光纤51Tbps。
"AI工厂的带宽挑战如何解决?" "光学互连在AI集群中的应用前景?" "多站点AI工厂的扩展之路揭秘!"
客服
商务合作
小程序
服务号
折叠