NVIDIA DGX Station
NVIDIA DGX Station
This product is intended for businesses, government and educational institutions only. Completing an order on the e-shop does not conclude a contract, but only submits a request for quotation.
The sale of this product is subject to export, sanction, and other legal restrictions. Before an order is accepted, it is necessary to verify the buyer, the end user, the country of delivery, and the intended use. An enquiry can only be accepted after successful authorisation from the supplier.
If the required conditions are not met, the request for quotation may be rejected or the order may be cancelled without any claim to the delivery of goods.
If you are interested in purchasing an AI solution or accelerator, you can also write to us at [email protected]. We will then prepare a quote and an individual contract for you. For more information, please see the terms and conditions (clauses 3.33 and 3.34).
NVIDIA DGX Station supercomputer
The most powerful AI supercomputer for your desk – data centre performance without the server room
The NVIDIA DGX Station is the ultimate deskside AI supercomputer for data scientists, AI researchers, and developers who need the power of an entire data centre right at their workstation. At its core is the NVIDIA GB300 Grace Blackwell Ultra superchip – a combination of the Blackwell Ultra GPU architecture with 252GB of HBM3e memory and a 72-core Grace Neoverse V2 ARM processor with 496GB of LPDDR5X memory, connected by NVLink-C2C technology with 900GB/s throughput. The result is a coherent memory space capable of holding and locally running models with up to 1 trillion parameters at up to 20 petaFLOPs in FP4 format – without cloud dependency, shared infrastructure, and with full control over sensitive data. NVIDIA MIG technology allows the GPU to be partitioned into up to seven isolated instances for the entire AI team to work concurrently, and connecting two DGX Stations via the NVIDIA ConnectX-8 SuperNIC with up to 800Gb/s bandwidth doubles the performance to over 1.4TB of coherent memory and 40 petaFLOPs of AI power. The pre-installed NVIDIA DGX OS on Ubuntu 24.04 LTS with CUDA-X libraries, TensorRT, NIM microservices, and NVIDIA AI Enterprise ensures an immediate start without lengthy configuration.
748GB (HBM3e + LPDDR5x)
Coherent memory
Blackwell GB300
Chip architecture
NVIDIA ConnectX
High-speed network card
NVIDIA AI Software Stack
Complete software stack
1600W
Power consumption
Key features of the NVIDIA DGX Station workstation
- NVIDIA GB300 Grace Blackwell Ultra superchip with Blackwell Ultra GPU (252GB HBM3e, 7.1TB/s) and 72-core Grace Neoverse V2 ARM processor (496GB LPDDR5X) connected by 900GB/s NVLink-C2C technology
- Up to 20 petaFLOPs of FP4 performance and 330 TOPS of Int8 performance for training, fine-tuning and inference of the most demanding AI models
- Coherent memory capable of loading and running models with up to 1 trillion parameters locally, without dependency on the cloud or an external server
- NVIDIA MIG technology to partition the GPU into up to 7 isolated instances – a powerful local cluster for the entire AI team in a single device
- Connection of two DGX Stations via ConnectX-8 SuperNIC (up to 800Gb/s) for over 40 petaFLOPs of performance and over 1.4TB of coherent memory
- Pre-installed NVIDIA DGX OS with CUDA-X, TensorRT, NIM microservices, NVIDIA AI Enterprise, PyTorch, and TensorFlow for an immediate start without configuration
GB300 Grace Blackwell Ultra superchip and 748GB of coherent memory
The heart of the NVIDIA DGX Station is the GB300 Grace Blackwell Ultra superchip – a unique combination of a Blackwell Ultra GPU with 252GB of HBM3e memory and 7.1TB/s throughput, and a 72-core Grace Neoverse V2 ARM processor with 496GB of LPDDR5X and 396GB/s throughput. Both parts of the superchip are connected by NVLink-C2C technology with a bandwidth of 900GB/s, which eliminates the traditional bottleneck between CPU and GPU memory and creates a single 748GB coherent memory space. The result is 20 petaFLOPs of FP4 performance and 330 TOPS of Int8 performance – right where an entire server room floor was once needed.
A supercomputer for the AI team – MIG and ConnectX-8 clustering
The NVIDIA DGX Station is not just a personal supercomputer, but a powerful local compute node for the entire AI team. NVIDIA MIG technology allows the GPU to be partitioned into up to seven fully isolated instances, each with its own dedicated memory, compute cores, and network connection, for multiple researchers to work concurrently without impacting each other. Connecting two DGX Stations via the NVIDIA ConnectX-8 SuperNIC with up to 800Gb/s bandwidth then doubles the performance to over 40 petaFLOPs and over 1.4TB of coherent memory to work with the largest available models – without the need to build an expensive server cluster.
Acceleration of all AI tasks in a compact body
The power of the Grace Blackwell architecture in a body that fits on any desk – the NVIDIA DGX Station is the ideal choice for developers, researchers, and data scientists who need full-fledged AI performance without compromise. Whether you're working on large language model inference, specialising pre-trained networks, or developing AI agents, the DGX Spark handles the full spectrum of AI tasks locally, quickly, and without reliance on cloud infrastructure.
Prototyping and development of AI applications without limits
The complete NVIDIA AI software stack provides developers with a full-fledged platform for creating AI models, AI agents, and AI-enhanced applications – all locally, without cloud dependency. NVIDIA DGX OS on Ubuntu 24.04 LTS comes pre-configured with CUDA-X libraries, Container Toolkit, Docker, and NVIDIA AI Enterprise for an immediate start to development. Once the solution is ready for production deployment, the DGX Station enables direct and seamless migration of workloads to the NVIDIA DGX Cloud or other NVIDIA-accelerated infrastructure – with the same tools, libraries, and pre-trained models.
Fine-tune AI models with billions of parameters right at your workstation
Use the NVIDIA DGX Station to fine-tune pre-trained models with billions of parameters right at your workstation. Training on your own data enables you to specialise AI models for specific industry data or use cases – without sending sensitive data to the cloud or paying for expensive GPU instances. The NemoClaw platform then ensures secure and isolated operation of autonomous AI agents with full governance control. The result is a model tailored precisely to your scenario, created locally, securely, and under your full control.
NVIDIA DGX Station delivers top-tier data science
The combination of 748GB of coherent memory and 20 petaFLOPS of performance makes the NVIDIA DGX Station the ideal workstation for demanding data analytics and machine learning projects. Extensive datasets, complex computational models, and demanding training processes that previously required a powerful cloud cluster or a dedicated server can now run right on your desk. Four M.2 Gen 5 slots, two 400Gb/s QSFP112 network interfaces, and a PCIe Gen 5 x16 ensure data throughput worthy of a research facility—fast, local, and without waiting for remote infrastructure.
Inference for models with up to 1 trillion parameters
The Blackwell Ultra architecture's Tensor Cores, with support for the FP4 format, achieve performance of up to 20 petaFLOPS. Combined with 748GB of coherent memory, they enable you to run inference on the largest available AI models with up to 1 trillion parameters directly on your desk. Test, validate, and deploy models locally and in real time—without cloud latency, shared infrastructure, or the risk of sensitive data leaks. NVIDIA NIM microservices then allow for the rapid deployment of optimised models as containers for production inference.
Application development for robotics, industry, and smart cities
NVIDIA DGX Station is an exceptional platform for developing robotic systems, smart city solutions, and computer vision applications. The pre-installed NVIDIA Isaac framework for robotics, Metropolis for video analytics, and Holoscan for real-time data processing allow developers to fully harness the power of the DGX Station and rapidly prototype and deploy applications—locally, without reliance on cloud infrastructure, and with full support from the NVIDIA ecosystem.
AI supercomputer performance
Up to 20 petaFLOPS for FP4 format inference
Up to 1 trillion parameters
Local inference without cloud dependency
Connecting two systems
Cluster for over 40 petaFLOPS of performance and 1.4TB of coherent memory
Ready to work immediately
Pre-installed NVIDIA AI software stack
Memory without compromise
748GB of unified CPU and GPU memory
Your own data
Fine-tune models with billions of parameters directly at your workplace
Network throughput
ConnectX-8 SuperNIC with bandwidth up to 800Gb/s
Secure local work
Full data control without sending to the cloudSpecifications
Use
Processor
GPU series/model
Graphics card
Motherboard
Hard Drive
Colour and design
PSU
Memory
Outputs
Package contents
AI
Security and management
NCNR (Non-Cancellable, Non-Returnable)
AI specifications
GPU configuration
| Code: | NRn2001nad |
| Warranty: | 24 months |
| Links: | Producer's Website: |