DGX-B200, Nvidia DGX System, B200/8x GPUs

The NVIDIA DGX-B200 is a flagship AI system engineered for data center-scale training, inference, and simulation. Integrating 8 x Blackwell GPUs with 1,440 GB total GPU memory and 64 TB/s HBM3e bandwidth, it delivers extreme throughput for large language models and complex AI workloads. A 14.4 TB/s NVLink GPU interconnect minimizes latency between GPUs, while dual Xeon host CPUs provide robust orchestration, making the DGX-B200 a turnkey platform for enterprise and cloud AI deployment.

Product Features Comparison

Model	DGX-B200	BE6S-PRI-M2-K9, Cisco BE6000S Appliance, Unified Communications, Multi-service integration, Paging Server, Collaboration	BH640-V2, Huawei BH640 V2 Blade Server, Intel Xeon E5-4600/24 DDR3 DIMMs/For BE6000H	HGX-B200
Brand	NVIDIA	Cisco	Huawei	NVIDIA
Processor Model	Dual Xeon host CPUs	n/a	Intel Xeon E5-4600 and E5-4600 v2 series processors	n/a
Memory Capacity	1,440 GB total GPU memory	n/a	24 DDR3 DIMMs	180 GB HBM3e per GPU
Drive Bay Configuration	n/a	Cisco ISR 2921 ISR chassis with integrated UCS E160D M2 blade server	2 × 2.5-inch SAS/SATA HDDs or SSDs	n/a
Expansion Slots	n/a	n/a	Two PCIe x8 mezzanine slots for GE, 4G FC, 8G FC, and 10GE mezzanine cards	n/a
Power Supply	n/a	n/a	12 V DC	n/a