NVIDIA B200 192GB Blackwell Tensor Core GPU
Brand: NVIDIA | Category: GPUs
SKU: 900-21110-0000-000 | Part #: 900-21110-0000-000 | MPN: 900-21110-0000-000
Contact for Pricing — Request a Quote
About the NVIDIA B200 192GB Blackwell Tensor Core GPU
NVIDIA B200 192GB Blackwell Tensor Core GPU
The NVIDIA B200 is the flagship Blackwell-architecture GPU, delivering the highest AI training throughput available in a single accelerator. With 192GB of HBM3e memory at 8TB/s bandwidth and a peak FP8 performance of 9 PFLOPS, the B200 is engineered for the world's largest language model training runs, multi-modal AI systems, and scientific simulations that cannot be practically executed on earlier GPU generations.
Blackwell vs Hopper: The Performance Leap
The B200 delivers 5× more FP8 throughput and 2× more memory capacity than the H100 SXM, enabling models with significantly more parameters (GPT-5 class and beyond) to train within a single node rather than requiring multi-node parallelism. The NVLink 5.0 interconnect at 1.8 TB/s per GPU enables all-reduce operations at a bandwidth that fundamentally changes the communication-to-compute ratio in distributed training.
Key Specifications
| Parameter | Value |
|---|---|
| Architecture | NVIDIA Blackwell (GB100) |
| Memory | 192GB HBM3e |
| Memory Bandwidth | 8 TB/s |
| FP8 Performance | 9 PFLOPS |
| FP16 (Tensor) | 4.5 PFLOPS |
| NVLink | NVLink 5.0 (1.8 TB/s) |
| Interconnect | NVSwitch 4.0 |
| TDP | 1,000W |
| Form Factor | SXM6 (DGX B200 baseboard) |
Deployment Notes
The B200 requires NVIDIA DGX B200 or OEM-designed HGX B200 baseboard and 1kW power delivery per GPU. Contact Omnixon's AI Infrastructure team for full system configuration, cooling requirements, and UAE/GCC delivery timeline.
Technical Specifications
| Brand | NVIDIA |
| Category | GPUs |
| SKU | 900-21110-0000-000 |
| Part Number | 900-21110-0000-000 |
| Condition | New |
| Capacity | 192GB |
| Form Factor | SXM |
| RAID Support | H100 |
| GPU Support | NVIDIA B200 |
| AI Optimized | Yes |
| Architecture | Blackwell |
| Max Memory | 192GB HBM3e |
| Memory Bandwidth | Up to 8 TB/s |
| FP4 Performance | Up to 9 PFLOPS |
| TDP | 1000W |
| Interconnect | NVLink 5.0 (1.8TB/s) |
| Form Factor | SXM |
Frequently Asked Questions about NVIDIA B200 192GB Blackwell Tensor Core GPU
What does the NVIDIA B200 192GB Blackwell Tensor Core GPU do?
The NVIDIA B200 192GB Blackwell Tensor Core GPU accelerates AI/ML training, inference, scientific HPC, and virtualization (vGPU) workloads. Typical deployments include LLM training clusters, computer-vision pipelines, financial risk modeling, and rendering farms.
What are the headline specs of the NVIDIA B200 192GB Blackwell Tensor Core GPU?
Key specifications for the NVIDIA B200 192GB Blackwell Tensor Core GPU: new condition; brand NVIDIA; architecture Blackwell; model B200; memory 192GB HBM3e; memory bandwidth Up to 8 TB/s; fp4 performance Up to 9 PFLOPS. Manufacturer part number 900-21110-0000-000. For the full datasheet with electrical, environmental, and compliance details, contact our pre-sales engineering team.
Is the NVIDIA B200 192GB Blackwell Tensor Core GPU compatible with my infrastructure?
The NVIDIA B200 192GB Blackwell Tensor Core GPU requires a PCIe Gen4 or Gen5 x16 slot, server power adequate for the card's TDP, and CUDA/ROCm driver support in your hypervisor or bare-metal OS. Sales engineering will confirm chassis fit (1U/2U/4U), PCIe lane count, and PSU headroom before quoting.