NVIDIA DGX B300 - Enterprise AI Training and Reasoning

NVIDIA DGX™ B300

NVIDIA DGX™ B300 brings NVIDIA Blackwell Ultra GPUs to enterprise AI training, inference, and reasoning. Its eight-GPU design combines 2.1 TB of GPU memory with NVIDIA ConnectX-8 networking to support demanding generative AI workloads. An air-cooled chassis supports NVIDIA MGX and traditional enterprise racks, while integrated software helps teams deploy and operate their AI infrastructure.

An AI Factory for the Era of Reasoning

Real-Time Powerhouse for Inference and Training

NVIDIA Blackwell Ultra GPUs give enterprises one DGX B300 platform for large language model training and inference. With 144 petaFLOPS of FP4 sparse inference performance, the system brings hyperscale AI capabilities to enterprise workloads.

Efficient and Sustainable Innovation

DGX B300 offers multiple power configurations to suit different data center requirements. Its design focuses on energy efficiency and AI performance per watt, helping enterprises make effective use of available power.

Revolutionary Infrastructure Standard

A redesigned DGX B300 chassis brings NVIDIA MGX rack compatibility to the DGX platform for the first time. Built for modern data centers, it helps enterprises adopt infrastructure designed for greater AI performance and efficiency.

Specifications

	NVIDIA DGX 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/FP6 Tensor Core: 72 PFLOPS
NVIDIA NVLink Switch System	2x NVLink switches
NVIDIA NVLink Bandwidth	14.4 TB/s aggregate
Power Consumption	14.5 kW (busbar) / 15.1 kW (PSU)
CPU	Intel® Xeon® 6776P processors
System Memory	2 TB, configurable to 4 TB
Networking	8x OSFP ports serving 8x single-port NVIDIA ConnectX-8 VPI Up to 800 Gb/s InfiniBand/Ethernet 2x dual-port QSFP112 NVIDIA BlueField-3 DPU Up to 400 Gb/s InfiniBand/Ethernet
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 NVIDIA Mission Control with NVIDIA Run:ai technology NVIDIA DGX OS Supports Red Hat Enterprise Linux, Rocky Linux, and Ubuntu
Rack Units (RU)	10U
Operating Temperature	10–35°C (50–95°F)
Enterprise Support	Three-year business-standard hardware and software support

Specifications: NVIDIA DGX B300 and the DGX B300 datasheet. Power figures follow the datasheet's busbar and PSU configurations; NVIDIA's product page lists approximately 14 kW. Performance figures distinguish sparse and dense operation.

NVIDIA Mission Control

NVIDIA Mission Control brings cluster deployment, workload scheduling, monitoring, and recovery into one operations platform. NVIDIA Run:ai technology helps orchestrate workloads, while NVIDIA AI Enterprise supplies software for developing and deploying AI. Together, these tools help teams manage DGX B300 infrastructure and keep GPU resources productive.

NVIDIA DGX Enterprise Support and Services

DGX B300 includes three-year business-standard hardware and software support. NVIDIA Enterprise Services also offers access to support, education, and infrastructure specialists to help teams deploy and operate their DGX systems.

Read the Datasheet

Experience the Benefits

Ensure Continuity

Keep your DGX platform operating at peak performance with various levels of enterprise-grade support and a deep connection with the product and engineering.

Achieve Faster ROI

Get faster results with onboarding, optimization, training, and certification services for your employees.

Customize Your AI Journey

Get personalized instructions and guidance and proactive recommendations for your data science practice and data center operations.

Servers/JBODs

WD

Public Test Drives

Customer Remote Access

Build Your AI Factory for Reasoning