|
戴尔IR7000配备直接液体冷却：每机架高达480kW，搭载GB200 NVL4和144个B200 GPU
2024年11月25日 17:55 上海
戴尔最新推出的IR7000系统通过直接液体冷却技术，实现每机架高达480kW的计算能力：支持灵活设计，可搭载NVIDIA GB200 NVL4或GB200 NVL72。
戴尔CEO迈克尔·戴尔在X平台发文：
“我们的新IR7000通过直接液体冷却技术支持每机架高达480kW的计算能力。这种灵活的设计支持任意组合的云计算、AI和高性能计算（HPC）。该系统可捕获每个机架内近100%的热量。”
戴尔为其新款IR7000服务器机架提供了三个示例配置
1. NVIDIA GB200 NVL4，配备144个B200 GPU（总计30万亿个晶体管）
2. NVIDIA GB200 NVL72，配备72个B200 GPU
3. 72个AMD EPYC第五代CPU节点，每机架最高可达27,000个核心
戴尔尖端的集成机架可扩展系统（IRSS）专为最密集的人工智能工作负载而设计，该公司定制了服务器机架，以最大限度地利用空间、提高能源效率并降低成本。
更多关于DELL IR7000系列的信息
DEL IR7000系列是一种面向下一代基础设施的机架系统，重新定义了高密度计算时代的数据中心。它为大规模AI工作负载设计，具备高核心密度和直接液体冷却功能。作为基于开放计算项目（OCP）标准的基础设施，IR7000支持多代和异构技术，并提供多种机架计算选项。
IR7000系列的关键特点：
- 21英寸的Orv-3机架基础设施：专为高功率设计（高TDP的GPU和CPU），采用密集计算和液体冷却技术。
- 集成供电母线和液冷分配管：简化机架背部维护和布线，实现标准化。
- 供电架：消除PSU/PDU布线的复杂性，同时支持机架内和机架行级冷却分配单元（CDU）。
戴尔IR7000系列的客户优势：
1. 无与伦比的简化：无需布线的液体和电力传输。
2. 可扩展性：随计算需求增长而扩展。
3. 快速部署：提供从数据中心评估到机架级集成的“白手套”服务。
4. 未来设计：每机架支持高达480kW。
5. 高效节能：集成液体冷却技术。
6. 高密度：具备最高的CPU和GPU密度。
7. 标准化：基于OCP标准的大规模AI和密集计算基础设施。
8. 灵活性：支持多代和异构技术（CPU、GPU以及CPU+GPU组合）。
9. 即用型机架部署：从数据中心评估到部署全程支持。
-end-
本文由“壹伴编辑器”提供技术支持
*版权声明：尊重原创，内容素材来源于网络，若涉及版权问题，请及时联系进行删除或合作事宜。
Dell AI Factory Offers New Cooling, High Density Compute and AI Storage
October 15, 2024 by staff
Round Rock, TEXAS – October 15, 2024: Dell Technologies (NYSE: DELL) introduced integrated rack-scalable systems, server, storage and data management innovations to the Dell AI Factory, for high density computing and AI workloads at scale.
The Dell Integrated Rack 7000 (IR7000) is designed to handle accelerated computing demands with high density, more sustainable power management and advanced cooling technologies. This Open Compute Project (OCP) standards-based rack is ideal for large-scale deployment and features a futureproof design for multigeneration and heterogenous technology environments.
“Today’s data centers can’t keep up with the demands of AI, requiring high density compute and liquid cooling innovations with modular, flexible and efficient designs,” said Arthur Lewis, president, Infrastructure Solutions Group, Dell Technologies. “These new systems deliver the performance needed for organizations to remain competitive in the fast-evolving AI landscape.”
Key features include:
Designed for density, the 21-inch Dell IR7000 is designed to support industry-leading CPU and GPU density.
Future-ready and efficient, the rack features wider, taller server sleds to accommodate the latest, larger CPU and GPU architectures. This rack was purpose built for liquid cooling natively, capable of cooling future deployments of up to 480KW, and is able to capture nearly 100% of heat created.
Engineered for greater choice and flexibility, this integrated rack offers support for both Dell and off-the-shelf networking.
Deployments are simple and energy-efficient with Dell Integrated Rack Scalable Systems (IRSS). IRSS delivers innovative rack-scale infrastructure optimized for AI workloads, making the setup process seamless and efficient with a fully integrated plug-and-play rack scale system.
Dell Technologies introduces AI-ready platforms designed for the Dell IR7000:
Part of the Dell AI Factory with NVIDIA, the Dell PowerEdge XE9712 offers high-performance, dense acceleration for LLM training and real-time inferencing of large-scale AI deployments. Designed for industry-leading GPU density with NVIDIA GB200 NVL72, this platform connects up to 36 NVIDIA Grace CPUs with 72 NVIDIA Blackwell GPUs in a rack-scale design. The 72 GPU NVLink domain acts as a single GPU for up to 30x faster real-time trillion-parameter LLM inferencing. The liquid cooled NVIDIA GB200 NVL72 is up to 25x more efficient than the air-cooled NVIDIA H100-powered systems.
The Dell PowerEdge M7725 provides high performance dense compute ideal for research, government, fintech and higher education environments. Designed to be deployed in the IR7000 rack, the Dell PowerEdge M7725 delivers more compute with improved serviceability scaling between 24K-27K cores per rack, with 64 or 72 two socket nodes, powered by 5th Gen AMD EPYC CPUs Front IO slots enables high speed IO connectivity and provides seamless connectivity for demanding applications. The server’s energy-efficient form factor allows for more sustainable deployments through both direct liquid cooling (DLC) to CPUs and air cooling via quick connect to the integrated rack.
Unstructured storage and data management innovations for the AI era
Dell Technologies unstructured data storage portfolio innovations improve AI application performance and deliver simplified global data management.
Dell PowerScale, the world’s first Ethernet storage certified for NVIDIA DGX SuperPOD, delivers new updates that enhance data management strategies, improve workload performance and offer greater support for AI workloads.1
Enhanced discoverability: Unlock data insights for faster smarter decision-making using PowerScale metadata and the Dell Data Lakehouse. A forthcoming Dell open-source document loader for NVIDIA NeMo services and RAG frameworks is designed to help customers improve data ingestion time and decrease compute and GPU cost.
Denser storage: Customers can fine tune their AI models by training them on larger datasets with new 61TB drives that increase capacity and efficiency while reducing data center storage footprint by half.2
Improved AI performance: AI workload performance is enhanced through front-end NVIDIA InfiniBand capabilities and 200GbE Ethernet adapter support that delivers up to 63% faster throughput.3
With new enhancements to the Dell Data Lakehouse data management platform, customers can save time and improve operations with new features like disaster recovery, automated schema discovery, comprehensive management APIs, and self-service full stack upgrades.
Customers can simplify their data-driven journey and quickly scale their AI and business use cases with Optimization Services for Data Cataloging and Implementation Services for Data Pipelines. These services increase accessibility to high-quality data through discovery, organization, automation and integration.
Dell Generative AI Solutions with Intel for modern workflows
As part of the Dell AI Factory, Dell Generative AI Solutions with Intel offers jointly engineered, tested and validated platforms for seamless AI deployment. Featuring the Dell PowerEdge XE9680 and Intel ? Gaudi 3 ? AI accelerators with Dell storage, networking, services and an open-source software stack, these preconfigured, flexible and high performing solutions support a range of GenAI use cases including content creation, digital assistants, design and data creation, code generation and more.
Availability
The Dell IR7000 will be globally available Q1 CY2025.
The Dell PowerEdge XE9712 is sampling for select customers now.
The Dell PowerEdge M7725 will be globally available Q1 CY2025.
Dell PowerScale updates will be available in Q4 CY2024.
Dell Data Lakehouse updates will be available in 1H CY2025.
Dell Generative AI Solutions with Intel will be available in Q4 CY2024.
Filed Under: Compute, CPUs, GPUs, FPGAs, HPC Hardware, HPC Software, Machine Learning, News
Tagged With: Dell, Dell AI Factory, Dell Technologies, nvidia
Dell IR7000 with Direct Liquid Cooling: up to 480kW per rack: GB200 NVL4 with 144 x B200 GPUs
Dell's new IR7000 supports up to 480kW per rack of computing power with Direct Liquid Cooling: flexible design with NVIDIA GB200 NVL4 or GB200 NVL72.
Dell IR7000 with Direct Liquid Cooling: up to 480kW per rack: GB200 NVL4 with 144 x B200 GPUs
Dell's new IR7000 supports up to 480kW per rack of computing power with Direct Liquid Cooling: flexible design with NVIDIA GB200 NVL4 or GB200 NVL72.
Published Nov 24, 2024 11:11 PM CST
TL;DR: Dell announced the IR7000 servers, supporting up to 480kW per rack with Direct Liquid Cooling, designed for Cloud, AI, and HPC. The IR7000 series features a scalable, sustainable infrastructure with high core density and advanced cooling.
Dell has just announced its new IR7000 servers that support up to 480kW per rack of computing power with Direct Liquid Cooling. Check it out:
Dell CEO Michael Dell posted on X saying: "Our new IR7000 supports up to 480kW per rack of computing power with Direct Liquid Cooling. The flexible design allows for any combination of Cloud, AI and HPC. The system is engineered to capture nearly 100% of the heat produced within each rack".
Popular Now: SpaceX launches Starlink's new direct-to-cell (DTC) technology, seamless global coverage
Dell has three example configurations of its new IR7000 server racks:
NVIDIA GB200 NVL4 with 144 B200 GPUs (30 trillion transistors)
NVIDIA GB200 NVL72 with 72 B200 GPUs
72 nodes of AMD EPYC 5th Gen CPU with up to 27,000 cores per rack
Dell's cutting-edge Integrated Rack Scalable Systems (IRSS) are designed for the most intensive AI workloads, with the company custom building the server racks to maximize space, enhance energy efficiency, and reduce costs.
Read more: Dell PowerEdge XE9712: NVIDIA GB200 NVL72-based AI GPU cluster
Integrated Rack 7000 (IR7000) series:
The IR7000 series to rack is a next-generation infrastructure that reimagines data centers for the new era of high-density computing. Designed for scalability, sustainability, and advanced cooling technologies, the IR7000 series to rack is optimized for large, rack-scale AI workloads with high core density and direct liquid cooling. As an Open Compute Project (OCP) Foundation standards-based infrastructure to rack, the IR7000 scales As an Open Compute Project (OCP) Foundation standards-based infrastructure, the IR7000 scales to support multigeneration and heterogeneous technology with diverse rack compute options.
The IR7000 series features a 21" Orv-3-based rack infrastructure with dense compute and liquid cooling for high TDP GPUs and CPUs. With integrated power busbars and DLC manifolds, the IR7000 series simplifies back-of-rack serviceability and cabling, paving the way for standardization. The series also includes power shelves that eliminate PSU/PDU cabling complexity and supports both in-rack and in-row CDUs.
Customer Benefits of the Dell IR7000 series:
Unparalleled simplicity: Cable-free liquid and power delivery
Scalable: Grow as your compute demand grows
Rapid deployment: Your entire HPC cluster at scale with a white-glove experience
Future-ready design: Support up to 480kW in each rack
Efficient: Integrated DLC for energy efficiency
Dense: Highest CPU and GPU density
Standardized: OCP standard-based infrastructure for large-scale AI & dense compute
Flexibility: Designed for multigeneration & heterogeneous technology (CPU, GPU & CPU+GPU)
Turnkey Rack Level Deployment: Unmatched service and support from datacenter assessment to rack scale integration
微信扫一扫
关注该公众号
关于我们
北京汉深流体技术有限公司是丹佛斯中国数据中心签约代理商。产品包括FD83全流量自锁球阀接头；液冷通用快速接头UQD & UQDB；OCP ORV3盲插快换接头BMQC；EHW194 EPDM液冷软管、电磁阀、压力和温度传感器及Manifold的生产。在国家数字经济、东数西算、双碳、新基建战略的交汇点，公司聚焦组建高素质、经验丰富的液冷工程师团队，为客户提供卓越的工程设计和强大的客户服务。
公司产品涵盖：丹佛斯液冷流体连接器、EPDM软管、电磁阀、压力和温度传感器及Manifold。
未来公司发展规划：数据中心液冷基础设施解决方案厂家，具备冷量分配单元（CDU）、二次侧管路（SFN）和Manifold的专业研发设计制造能力。
- 针对机架式服务器中Manifold/节点、CDU/主回路等应用场景，提供不同口径及锁紧方式的手动和全自动快速连接器。
- 针对高可用和高密度要求的刀片式机架，可提供带浮动、自动校正不对中误差的盲插连接器。以实现狭小空间的精准对接。
- 基于OCP标准全新打造的液冷通用快速接头UQD & UQDB ;OCP ORV3盲插快换接头BMQC , 支持全球范围内的大批量交付。
|
