Brand: NVIDIA | Category: GPUs
SKU: 900-23001-0020-000 | Part #: 900-23001-0020-000 | MPN: 900-23001-0020-000
Contact for Pricing — Request a Quote
Large-language model training and AI inference workloads demand the NVIDIA Blackwell architecture, and the B200 PCIe 192GB HBM3e GPU delivers exactly that foundation. This datacenter accelerator combines 192 GB of HBM3e memory with an 8 TB/s bandwidth ceiling, enabling enterprises to train transformer models at scale without memory bottlenecks. The B200 (part number 900-23001-0020-000) achieves 9 PFLOPS in FP8 tensor performance and 4.5 PFLOPS in FP16, making it purpose-built for both dense inference and multi-GPU training clusters. Its 2nd Generation Transformer Engine and 5th Generation NVLink support further reduce latency in large-batch operations, while PCIe Gen6 x16 connectivity integrates cleanly into standard enterprise infrastructure.
AI infrastructure teams, machine learning operations managers, and platform engineers procure the B200 when conventional GPUs no longer meet their throughput demands. Organizations running proprietary large language models, fine-tuning foundation models, or deploying real-time inference services at hyperscale choose NVIDIA accelerators for their software ecosystem and proven reliability. The dedicated RAS engine and ECC support on HBM3e memory address mission-critical uptime requirements in production datacenters. With a 700 W TDP and passive cooling design, the B200 integrates into existing modular liquid-cooled or hot-aisle infrastructure without proprietary retrofit investments.
Omnixon Global stocks enterprise-grade accelerators for regional AI and HPC deployments across the Middle East and beyond. Our procurement team holds direct relationships with NVIDIA and ensures fast verification, customs clearance, and logistics to your Dubai or regional facility. If your organization is evaluating the B200 for next-generation training clusters or inference platforms, contact Omnixon Global to request a formal quotation and technical specification review for part number 900-23001-0020-000.
| Brand | NVIDIA |
| Category | GPUs |
| SKU | 900-23001-0020-000 |
| Part Number | 900-23001-0020-000 |
| Condition | New |
| Manufacturer Part Number | 900-23001-0020-000 |
| GPU Architecture | NVIDIA Blackwell |
| Form Factor | PCIe |
| HBM Capacity | 192 GB HBM3e |
| Number of HBM Stacks | 8 |
| Memory Bandwidth | 8 TB/s |
| Interface | PCIe Gen6 x16 |
| TDP (Total Board Power) | 700 W |
| FP8 Tensor Core Performance | 9 PFLOPS |
| FP16 Tensor Core Performance | 4.5 PFLOPS |
| FP32 Performance | 1.8 PFLOPS |
| FP64 Performance | 900 TFLOPS |
| Transformer Engine | 2nd Generation |
| NVLink Generation | 5th Generation NVLink |
| RAS Engine | Yes — dedicated Reliability, Availability, and Serviceability Engine |
| Error Correction | ECC supported on HBM3e memory |
| CUDA Compute Capability | 10.0 |
| Cooling | Passive (datacenter airflow required) |
| Supported Precision Formats | FP4, FP8, FP16, BF16, TF32, FP32, FP64, INT8 |
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.
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.
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.
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.