{"id":5386,"date":"2024-07-18T08:29:32","date_gmt":"2024-07-18T08:29:32","guid":{"rendered":"https:\/\/openzeka.com\/en\/?page_id=5386"},"modified":"2024-12-13T07:04:07","modified_gmt":"2024-12-13T07:04:07","slug":"dgx-b200","status":"publish","type":"page","link":"https:\/\/openzeka.com\/en\/dgx-b200\/","title":{"rendered":"DGX B200"},"content":{"rendered":"

NVIDIA DGX B200<\/h1><\/div>

The foundation for your AI center of excellence.<\/p>\n<\/div><\/div><\/div><\/div><\/div>

NVIDIA DGX SuperPOD<\/h1><\/div>

Purpose-Built for the Unique Demands of AI<\/p>\n<\/div><\/div><\/div><\/div><\/div>

NVIDIA DGX SuperPOD<\/h1><\/div>

Purpose-Built for the Unique Demands of AI<\/p>\n<\/div><\/div><\/div><\/div><\/div>

Revolutionary Performance Backed by Evolutionary Innovation<\/h1><\/div>

NVIDIA DGX B200 is an unified AI platform for develop-to-deploy pipelines for businesses of any size at any stage in their AI journey. Equipped with eight NVIDIA Blackwell GPUs interconnected with fifth-generation NVIDIA NVLink, DGX B200 delivers leading-edge performance, offering 3X the training performance and 15X the inference performance of previous generations. Leveraging the NVIDIA Blackwell GPU architecture, DGX B200 can handle diverse workloads\u2014including large language models, recommender systems, and chatbots\u2014making it ideal for businesses looking to accelerate their AI transformation.<\/p>\n<\/div><\/div><\/div><\/div><\/div>

A Unified AI Platform<\/h1><\/div><\/div><\/div>
<\/span><\/div>

One Platform for Develop-to-Deploy Pipelines<\/h1><\/div>

Enterprises require massive amounts of compute power to handle complex AI datasets at every stage of the AI pipeline, from training to fine-tuning to inference. With NVIDIA DGX B200, enterprises can arm their developers with a single platform built to accelerate their workflows.<\/p>\n<\/div><\/div><\/div>

<\/span><\/div>

Powerhouse of AI Peformance<\/h1><\/div>

Powered by the NVIDIA Blackwell architecture\u2019s advancements in computing, DGX B200 delivers 3X the training performance and 15X the inference performance of DGX H100. As the foundation of NVIDIA DGX BasePOD and NVIDIA DGX SuperPOD, DGX B200 delivers leading-edge performance for any workload.<\/p>\n<\/div><\/div><\/div>

<\/span><\/div>

Proven Infrastructure Standard<\/h1><\/div>

DGX B200 is a fully optimized hardware and software platform that includes the complete NVIDIA AI software stack, including NVIDIA Base Command and NVIDIA AI Enterprise software, a rich ecosystem of third-party support, and access to expert advice from NVIDIA professional services.<\/p>\n<\/div><\/div><\/div><\/div><\/div>

Next-Generation Performance Powered by DGX B200<\/h1><\/div><\/div><\/div>

Real Time Large Language Model Inference<\/strong><\/h1><\/div>
<\/span><\/div><\/div><\/div>

Supercharged AI Training Performance<\/strong><\/h1><\/div>
<\/span><\/div><\/div><\/div>

Projected performance subject to change. Token-to-token latency (TTL) = 50ms real time, first token latency (FTL) = 5s, input sequence length = 32,768, output sequence length = 1,028, 8x eight-way DGX H100 GPUs air-cooled vs. 1x eight-way DGX B200 air-cooled, per GPU performance comparison\u200b.<\/p>\n<\/div><\/div><\/div>

Projected performance subject to change. 32,768 GPU scale, 4,096x eight-way DGX H100 air-cooled cluster: 400G IB network, 4,096x 8-way DGX B200 air-cooled cluster: 400G IB network.<\/p>\n<\/div><\/div><\/div><\/div><\/div>

NVIDIA DGX B200 Specifications<\/h1><\/div><\/div><\/div>
\n
\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
GPU<\/th>\n8x NVIDIA Blackwell GPUs<\/th>\n<\/tr>\n<\/thead>\n
GPU Memory<\/td>\n1,440GB total GPU memory<\/td>\n<\/tr>\n
Performance<\/td>\n72 petaFLOPS training and 144 petaFLOPS inference<\/td>\n<\/tr>\n
Power Consumption<\/td>\n~14.3kW max<\/td>\n<\/tr>\n
CPU<\/td>\n2 Intel\u00ae Xeon Platinum 8570 Processors
\n112 Cores total, 2.1 GHz (Base),
\n4 GHz (Max Boost)<\/td>\n<\/tr>\n
System Memory<\/td>\n Up to 4TB<\/td>\n<\/tr>\n
Networking<\/td>\n4x OSFP ports serving 8x single-port NVIDIA ConnectX-7 VPI
\nUp to 400Gb\/s InfiniBand\/Ethernet
\n2x dual-port QSFP112 NVIDIA BlueField-3 DPU
\nUp to 400Gb\/s InfiniBand\/Ethernet<\/td>\n<\/tr>\n
Management Network<\/td>\n10Gb\/s onboard NIC with RJ45
\n100Gb\/s dual-port ethernet NIC
\nHost baseboard management controller (BMC) with RJ45<\/td>\n<\/tr>\n
Storage<\/td>\nOS: 2x 1.9TB NVMe M.2
\nInternal storage: 8x 3.84TB NVMe U.2<\/td>\n<\/tr>\n
Software<\/td>\nNVIDIA AI Enterprise: Optimized AI Software
\nNVIDIA Base Command: Orchestration, Scheduling, and Cluster Management
\nDGX OS \/ Ubuntu: Operating system<\/td>\n<\/tr>\n
Rack Units (RU)<\/td>\n10 RU<\/td>\n<\/tr>\n
System Dimensions<\/td>\nHeight: 17.5in (444mm)
\nWidth: 19.0in (482.2mm)
\nLength: 35.3in (897.1mm<\/td>\n<\/tr>\n
Operating Temperature<\/td>\n5\u201330\u00b0C (41\u201386\u00b0F)<\/td>\n<\/tr>\n
Enterprise Support<\/td>\nThree-year Enterprise Business-Standard Support for hardware and software
\n24\/7 Enterprise Support portal access
\nLive agent support during local business hours<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/div><\/div><\/div><\/div>

Built from the ground up for enterprise AI, the NVIDIA DGX platform combines the best of NVIDIA software, infrastructure, and expertise in a modern, unified AI development solution, available in the cloud or on-premises.<\/p>\n<\/div>

NVIDIA  DGX Platform >><\/a><\/p>\n<\/div><\/div><\/div><\/div><\/div>

For More Information<\/h1><\/div>