NVIDIA DGX B200: 8x Blackwell AI System

Home
Accelerators
Nvidia
NVIDIA DGX B200

Blackwell Architecture

Introduction

Revolutionary Performance Backed by Evolutionary Innovation

The NVIDIA DGX B200 is a unified AI platform for develop-to-deploy pipelines, built for businesses of any size at any stage of their AI journey. Equipped with eight NVIDIA Blackwell GPUs interconnected by fifth-generation NVLink, the DGX B200 delivers 3X the training performance and 15X the inference performance of previous-generation DGX systems. From large language models and recommender systems to chatbots and scientific simulation, the DGX B200 handles it all in a single 10U system.

AI Infrastructure for Any Workload

The DGX B200 combines eight Blackwell GPUs with 1.44 TB of HBM3e memory and 64 TB/s of aggregate bandwidth in a fully integrated system. Pre-loaded with the complete NVIDIA AI software stack, it accelerates every stage of the AI pipeline from data preparation through production inference.

Read the Datasheet

Benefits

A Unified AI Factory

Platform for the Full Pipeline

Enterprises need massive compute at every stage of the AI pipeline. The DGX B200 provides a single system for training, fine-tuning, and inference, eliminating the need to manage separate infrastructure for each workload stage.

15X

Inference Performance vs. DGX H100

Powered by Blackwell architecture advancements, the DGX B200 delivers 3X training performance and up to 15X inference performance over DGX H100. It serves as the foundation for DGX BasePOD and DGX SuperPOD deployments.

Full

Stack Software Included

A fully optimized hardware and software platform including NVIDIA Mission Control, NVIDIA AI Enterprise, a rich ecosystem of third-party support, and access to expert advice from NVIDIA professional services.

Performance

Next-Generation Performance

Real-Time Large Language Model Inference

The DGX B200 delivers up to 15X higher real-time inference throughput per GPU compared to DGX H100 on trillion-parameter models (GPT-MoE-1BT). With 1.44 TB of HBM3e memory and 14.4 TB/s of NVLink bandwidth, the system can serve massive models with the low latency that real-time applications demand.

View Performance Benchmarks

Supercharged AI Training Performance

At data center scale with 32,768 GPUs, DGX B200 clusters deliver dramatically faster training on trillion-parameter models compared to equivalent DGX H100 clusters. Fifth-generation NVLink at 1.8 TB/s per GPU enables all eight Blackwell GPUs to work in unison, while 400 Gb/s InfiniBand networking scales performance efficiently across nodes.

Learn About Blackwell Architecture

Server Compatibility

DGX B200 System & B200 GPU Servers

NVIDIA DGX B200 Complete System

The DGX B200 is a fully integrated 10U AI supercomputer with 8× Blackwell GPUs, 1.44 TB HBM3e, dual Intel Xeon Platinum CPUs, NVLink/NVSwitch interconnect, 400 Gb/s networking, 30+ TB NVMe storage, and the complete NVIDIA AI software stack. ServerMonkey can source and deliver DGX B200 systems for your AI factory.

Request a DGX B200 Quote

Dell PowerEdge Servers with B200 GPUs

For customers who prefer to deploy Blackwell GPUs within their existing Dell infrastructure.

Dell PowerEdge XE9780

Air-Cooled

Supports 8× B200 SXM GPUs (air-cooled)
17th Gen, purpose-built for large-scale AI

View Server

Dell PowerEdge XE9685L

Direct Liquid-Cooled

Supports 8× B200 SXM GPUs (DLC)
16th Gen, maximum density and efficiency

View Server

Dell PowerEdge XE9680L

Direct Liquid-Cooled

Supports 8× B200 SXM GPUs (DLC)
16th Gen, built for dense AI workloads

View Server

View Full GPU Compatibility Matrix

Specifications

NVIDIA DGX B200

Specification	DGX B200
GPU	8× NVIDIA Blackwell GPUs
GPU Memory	1,440 GB total HBM3e 64 TB/s aggregate bandwidth
FP4 Tensor Core	144 PFLOPS* 72 PFLOPS**
FP8 Tensor Core	72 PFLOPS*
NVIDIA NVSwitch	2×
NVLink Bandwidth	14.4 TB/s aggregate
CPU	2× Intel Xeon Platinum 8570 112 cores total, 2.1 GHz base / 4.0 GHz boost
System Memory	2 TB (configurable to 4 TB)
Networking	4× OSFP ports (8× ConnectX-7 VPI, up to 400 Gb/s) 2× dual-port QSFP112 BlueField-3 DPU (up to 400 Gb/s)
Management Network	10 Gb/s onboard NIC (RJ45) 100 Gb/s dual-port Ethernet NIC Host BMC with RJ45
Storage	OS: 2× 1.9 TB NVMe M.2 Internal: 8× 3.84 TB NVMe U.2
System Power	~14.3 kW max
Form Factor	10 RU
Dimensions	17.5" H × 19.0" W × 35.3" L (444 × 482.2 × 897.1 mm)
Operating Temperature	10–35°C (50–90°F)
Enterprise Support	3-year Enterprise Business-Standard (hardware + software)

* Sparse performance. ** Dense performance is ½ of sparse spec shown.

View Datasheet DGX Systems Documentation

Ready to Build Your AI Factory?

Get pricing, availability, and configuration details from the ServerMonkey team.

Request a Quote

NVIDIA DGX B200

Revolutionary Performance Backed by Evolutionary Innovation

AI Infrastructure for Any Workload

A Unified AI Factory

Next-Generation Performance

Real-Time Large Language Model Inference

Supercharged AI Training Performance

DGX B200 System & B200 GPU Servers

NVIDIA DGX B200 Complete System

Dell PowerEdge Servers with B200 GPUs

Dell PowerEdge XE9780

Dell PowerEdge XE9685L

Dell PowerEdge XE9680L

NVIDIA DGX B200

Ready to Build Your AI Factory?

What are you looking for?

Get to Know Us

Resources

What We Do

Customer Service

Follow Us