NVIDIA DGX Station GB300

A Grace Blackwell AI supercomputer on your desk.

NVIDIA DGX Station GB300 provides datacentre-level AI performance in a desktop device, delivering up to 20 PetaFLOPS performance. With the preinstalled NVIDIA AI software stack and 748GB of combined memory, you can develop AI models and data science projects locally and seamlessly deploy to the datacentre or cloud.

Unlike the DGX Spark, the DGX Station GB300 is available in different models from multiple brands such as Asus and MSI.

VIEW MODELS Let's Chat

Diagram showing NVIDIA Grace Blackwell AI supercomputer architecture

FEATURES

NVIDIA GPU, CPU, Networking, and AI Software Technologies

NVIDIA GB300 Superchip

The cutting-edge SoC architecture delivers outstanding performance with a combination of Blackwell Ultra GPU and Arm CPU cores.

NVIDIA NVLink-C2C Interconnect

A high-bandwidth connection between the GPU and CPU, using less power and die space than the PCIe bus.

748GB of coherent memory

The GPU is supported by 252GB of HMB3e memory with 7.1TB/s bandwidth; while the CPU is equipped with 496GB of LPDDR5X memory with 396GB/s bandwidth.

NVIDIA ConnectX Networking

Support for up to 800Gb/s QSFP InfiniBand network ports for ultra-fast data ingest or connecting up to two DGX Station GB300’s together in a cluster.

MIG

Multi-Instance GPU (MIG) fully isolates at the hardware level allowing memory, cache and cores to be partitioned into as many as seven independent instances, giving multiple users access to GPU acceleration.

NVIDIA AI Software Stack

A complete AI software stack including access to NVIDIA NIM microservices, Blueprints and AI Enterprise.*

*NVIDIA AI Enterprise licence available for purchase as an upgrade.

Maximise Your DGX Station GB300 With Production Ready Software

NVIDIA AI Enterprise, available to purchase as an upgrade with Station GB300, is a suite of frameworks and libraries, including NVIDIA NIM and NeMo microservices, that accelerate the development and deployment of AI projects. Leveraging pre-configured frameworks removes many of the manual tasks and complexity associated with software development, as each framework is tried, tested and optimised by NVIDIA. The less time spent developing, the greater the ROI on your AI hardware and data science investments.

Rather than trying to assemble thousands of co-dependent libraries and APIs from different authors when building your own AI applications, NVIDIA AI Enterprise removes this pain point by providing a full agentic AI software stack.

Diagram showing NVIDIA inference models and software workflow

NVIDIA DGX Station GB300 Workloads

Developing AI Models

The DGX Station GB300 is empowered by the same software stack as other NVIDIA AI appliances, making it easy to transfer your newly-developed AI model onto more powerful DGX, HGX, EGX and MGX systems, whether they’re in your own datacentre or in the cloud.

Running inference workloads on NVIDIA DGX systems

Inference Models

The growing size and complexity of generative AI models requires systems with large amounts of memory and fast networking. The DGX Station GB300 is the ultimate office-based system for AI developers, researchers and data science teams.

Data Science

DGX Station GB300 makes it easy to transfer your data analytics using the NVIDIA RAPIDS framework onto more powerful DGX, HGX, EGX and MGX systems, whether they’re in your own datacentre or in the cloud.

Team Cloud

The DGX Station GB300 is powerful enough to run workloads for up to seven members of your team simultaneously, negating the need to book time on centralised resources and upload data to and from such systems.

Buying the DGX Station GB300

The DGX Station GB300 is available in different models from multiple brands, configured by our in-house expert engineers. Explore the range in the tabs below.

Asus ExpertCenter Pro ET900N G3

This is the Asus implementation of the DGX Station GB300. It is a desktop-PC sized system based around the NVIDIA GB300 Superchip.

Key Specifications
GPU	NVIDIA Blackwell Ultra
GPU memory	252GB HBM3e
CPU	NVIDIA Grace‑72 Core Neoverse V2
System memory	496GB LPDDR5X
System drives SSDs	2x 2TB M.2 NVMe PCIe 5
Project drives SSDs	2x 2TB M.2 NVMe PCIe 5
Networking	2× 400Gb/s QSFP InfiniBand 1× 10Gb/s Ethernet 1× 1Gb/s Remote Management
Front Ports	2× USB 10Gb/s Type‑C 2× USB 10Gb/s 1× USB 2.0 1× 3.5mm minijack headphone 1× 3.5mm minijack microphone
Rear Ports	4× USB 10Gb/s (4× Type‑A) 1× USB Micro B (for BMC) 3× 3.5mm minijack audio 1× mini DisplayPort
Operating System	NVIDIA DGX
Power Supply	1600W 80PLUS Titanium
Dimensions	230 × 584 × 565mm

Configurable Options
Graphics card for display	NVIDIA RTX PRO 4000 Blackwell SFF or 2000 Blackwell
Networking	WiFi 7

MSI XpertStation WS300

This is the MSI implementation of the DGX Station GB300. It is a desktop-PC sized system based around the NVIDIA GB300 Superchip.

Key Specifications
GPU	NVIDIA Blackwell Ultra
GPU memory	252GB HBM3e
CPU	NVIDIA Grace‑72 Core Neoverse V2
System memory	496GB LPDDR5X
System drives SSDs	2x 2TB M.2 NVMe PCIe 5
Project drives SSDs	2x 2TB M.2 NVMe PCIe 5
Networking	2× 400Gb/s QSFP InfiniBand 1× 10Gb/s Ethernet 1× 1Gb/s Remote Management
Front Ports	2x USB Type-C 2x USB Type-A 1× USB 2.0 1× 3.5mm minijack headphone 1× 3.5mm minijack microphone
Rear Ports	4× USB 10Gb/s (4× Type‑A) 1× USB Micro B (for BMC) 3× 3.5mm minijack audio 1× mini DisplayPort
Operating System	NVIDIA DGX
Power Supply	1600W 80PLUS Titanium
Dimensions	248 x 528 x 583mm

Configurable Options
Graphics card for display	NVIDIA RTX PRO 4000 Blackwell SFF or 2000 Blackwell
Networking	WiFi 7

Frequently Asked Questions

How do I buy an NVIDIA DGX Station GB300?

The DGX Station GB300 is a reference platform from NVIDIA and is not available for sell. Scan sells a variety of systems based on this design from brands such as Asus and MSI. Our expert engineers can help configure and build a complete system for you based on one of these designs.

How does the NVIDIA DGX Station GB300 compare to the DGX Spark?

The DGX Station GB300 is designed for developing huge AI models and complex data science projects for individual users or teams of up to seven people. In contrast, the DGX Spark is designed for developing developing and fine-tuning generative AI and reasoning AI models by a single user. Use this link to find out more about the DGX SPARK.

01204 474210

NVIDIA DGX Station GB300 – Desktop AI Supercomputer for Generative & Reasoning Models