NVIDIA DGX B200 8x B200 SXM6 1.4TB System

NVIDIA DGX B200 8x B200 SXM6 1.4TB System

Brand: NVIDIA | Category: GPUs

SKU: 920-23681-2530-000 | Part #: 920-23681-2530-000 | MPN: 920-23681-2530-000

Contact for Pricing — Request a Quote

Request a Quote Contact Us

About the NVIDIA DGX B200 8x B200 SXM6 1.4TB System

The NVIDIA DGX B200 is a purpose-built AI supercomputing system integrating eight NVIDIA B200 SXM6 Tensor Core GPUs interconnected via fifth-generation NVLink, delivering unified high-bandwidth GPU-to-GPU communication across the full system. Built on NVIDIA's Blackwell architecture, the DGX B200 provides a combined 1.4 TB of HBM3e GPU memory across the eight GPUs, enabling organizations to train and run the largest foundation models, multi-modal networks, and scientific simulation workloads without data fragmentation across memory boundaries. The system ships with NVIDIA's full DGX software stack, including DGX OS, NVIDIA AI Enterprise-compatible tooling, and deep integration with NVIDIA Base Command for cluster orchestration.

The Blackwell B200 GPU generation introduces a second-generation Transformer Engine, a new Decompression Engine accelerating database query analytics, and support for FP4 and FP6 precision formats in addition to FP8, TF32, BF16, and FP64 workloads. Each B200 SXM6 GPU delivers substantial improvements in dense and sparse AI inference throughput compared to the prior Hopper generation, enabling organizations to serve significantly larger language models at production latency targets. NVLink Switch integration within the DGX B200 chassis provides 1.8 TB/s of all-to-all GPU bandwidth across all eight GPUs, removing inter-GPU bottlenecks that limit scaling in PCIe-attached configurations.

Designed for enterprise datacenter deployment, the DGX B200 supports integration into multi-node SuperPOD reference architectures via NVIDIA Quantum-2 InfiniBand or Spectrum-X Ethernet networking fabrics. The system is intended for organizations requiring on-premises AI infrastructure for training, fine-tuning, and high-throughput inference of large-scale models, as well as for HPC and computational fluid dynamics workloads that demand sustained FP64 double-precision performance. Omnixon Global supplies this system to enterprise and government buyers across the UAE, GCC, EMEA, and APAC regions.

Ideal for

  • Large language model pre-training and full fine-tuning of models with hundreds of billions of parameters on-premises
  • High-throughput generative AI inference serving for enterprise-grade LLM and multimodal model deployments
  • Retrieval-augmented generation (RAG) pipeline acceleration combining dense vector search with transformer inference
  • Scientific simulation and high-performance computing workloads requiring sustained FP64 double-precision throughput
  • Sovereign AI infrastructure buildout for government and regulated-industry organizations requiring data residency compliance
  • Multi-node DGX SuperPOD cluster nodes for distributed training across hundreds of GPUs via NVLink and InfiniBand fabrics

Technical specifications

ManufacturerNVIDIA
Manufacturer Part Number920-23681-2530-000
Product NameNVIDIA DGX B200 8x B200 SXM6 1.4TB System
GPU ArchitectureNVIDIA Blackwell
Number of GPUs8x NVIDIA B200 SXM6
Total GPU Memory1.4 TB HBM3e
GPU Memory per GPU192 GB HBM3e
GPU InterconnectNVLink 5th Generation, 1.8 TB/s all-to-all GPU bandwidth
NVLink Bandwidth (Total Bidirectional)1.8 TB/s
Supported Precision FormatsFP4, FP6, FP8, FP16, BF16, TF32, FP64
System Memory2.3 TB system DRAM
Storage60 TB NVMe SSD raw capacity
Network Interfaces8x NVIDIA ConnectX-7 VPI (400Gb/s InfiniBand / 400GbE per GPU), 2x additional OOB management ports
Form FactorRackmount, 10U
Operating SystemUbuntu Linux-based DGX OS
Software StackNVIDIA DGX Software Stack, NVIDIA AI Enterprise-compatible
Power SupplyRedundant 10.2 kW
CoolingLiquid-cooled (direct liquid cooling)

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.

Technical Specifications

BrandNVIDIA
CategoryGPUs
SKU920-23681-2530-000
Part Number920-23681-2530-000
ConditionNew
Manufacturer Part Number920-23681-2530-000
Product NameNVIDIA DGX B200 8x B200 SXM6 1.4TB System
GPU ArchitectureNVIDIA Blackwell
Number of GPUs8x NVIDIA B200 SXM6
Total GPU Memory1.4 TB HBM3e
GPU Memory per GPU192 GB HBM3e
GPU InterconnectNVLink 5th Generation, 1.8 TB/s all-to-all GPU bandwidth
NVLink Bandwidth (Total Bidirectional)1.8 TB/s
Supported Precision FormatsFP4, FP6, FP8, FP16, BF16, TF32, FP64
System Memory2.3 TB system DRAM
Storage60 TB NVMe SSD raw capacity
Network Interfaces8x NVIDIA ConnectX-7 VPI (400Gb/s InfiniBand / 400GbE per GPU), 2x additional OOB management ports
Form FactorRackmount, 10U
Operating SystemUbuntu Linux-based DGX OS
Software StackNVIDIA DGX Software Stack, NVIDIA AI Enterprise-compatible
Power SupplyRedundant 10.2 kW
CoolingLiquid-cooled (direct liquid cooling)

Frequently Asked Questions about NVIDIA DGX B200 8x B200 SXM6 1.4TB System

What server platforms accept the NVIDIA DGX B200 8x B200 SXM6 1.4TB System?

Reference servers include Dell PowerEdge XE9680 / XE9712, HPE Cray XD670, Lenovo ThinkSystem SR685a / SR675 V3, Supermicro AS-A21GE / SYS-821GE, Gigabyte G593 / G894, ASUS ESC. Share your target platform in the RFQ and we will confirm chassis-to-GPU compatibility and recommended NIC pairing.

How long is the lead time on AI GPUs?

Highly model-dependent. L40S / RTX-class: typically 3-6 weeks. H100/H200/B200 in SXM form factor: 12-16 weeks for whole-platform allocations. We quote genuine-channel ETAs only — no grey-market promises.

Do you supply matched networking (Quantum InfiniBand / Spectrum-X)?

Yes — Omnixon stocks the full NVIDIA networking lineup (Quantum-2 / Quantum-X InfiniBand, Spectrum-X Ethernet, ConnectX NICs, BlueField DPUs) so we can quote a complete training-cluster BOM, not just the GPUs.

Can you help with NVIDIA AI Enterprise licensing?

Yes. We hold genuine-channels for NVIDIA AI Enterprise software subscriptions. Add it to your RFQ and we quote node-aligned licensing along with the hardware.