NVIDIA® B300
The NVIDIA Blackwell Ultra architecture redefines AI reasoning with breakthrough performance for the age of generative AI.
Powered by NVIDIA Blackwell Ultra architecture, the DGX B300 is the AI factory foundation for AI reasoning, delivering 70 petaFLOPS training and 192 petaFLOPS inference performance — purpose-built to meet the computational demands of generative AI and large language models.
With B300 you get:
The NVIDIA B300, powered by Blackwell Ultra architecture, delivers unprecedented AI reasoning capabilities — enabling real-time inference with full intelligence at scale for complex reasoning tasks, agentic AI, and generative AI workloads.
With 70 petaFLOPS training performance and advanced memory capacity, the B300 dramatically accelerates LLM and foundation model training, reducing time-to-train for the world's largest AI models.
Powered by NVIDIA Blackwell Ultra architecture's advancements in parallel computing, NVIDIA B300 delivers transformative performance for AI-driven scientific discovery, digital twins, and complex HPC simulations.
| Form Factor | B300 | ||
|---|---|---|---|
| GPU | 8x NVIDIA Blackwell Ultra SXM | ||
| GPU Memory | 2.1 TB total, 62 TB/s HBM3e bandwidth | ||
| Performance | FP4 Tensor Core: 144 PFLOPS | 108 PFLOPS* FP8 Tensor Core: 72 PFLOPS** |
||
| Power Consumption | ~14 kW | ||
| CPU | Intel® Xeon® 6776P Processors | ||
| NVIDIA NVLink Switch System | 2x | ||
| NVIDIA NVLink Bandwidth | 14.4 TB/s aggregate bandwidth | ||
| Networking |
8x OSFP ports serving 8x single-port NVIDIA ConnectX-8 VPI 2x dual-port QSFP112 NVIDIA BlueField-3 DPU |
||
| Management Network | 1GbE onboard NIC with RJ45 1GbE RJ45 Host baseboard management controller (BMC) |
||
| Storage | OS: 2x 1.9 TB NVMe M.2 Internal storage: 8x 3.84 TB NVMe E1.S |
||
| Software |
NVIDIA AI Enterprise—Optimized AI software NVIDIA Mission Control—AI data center operations and orchestration with NVIDIA Run:ai technology NVIDIA DGX OS—Operating system |
||
| Rack Units (RU) | 10U | ||
| Enterprise Support | Three-year business-standard hardware and software support | ||
|
*Specification shown in sparse | dense **Specification in sparse. Dense is ½ sparse spec shown |
|||