Nvidia finally unveiled the DGX Station workstation, based on the GB300 Grace Blackwell and equipped with almost a petabyte of LPDDR5X memory

New Workstation from Nvidia – DGX Station

*Official Announcement*

Nvidia announced the release of the DGX Station—a compact yet powerful station first seen at GTC 2025. The system is designed for software developers, researchers, and big‑data specialists who need compute power beyond that of the smaller DGX Spark model.

1. Key Hardware Components
The chassis looks… Accelerator Nvidia GB300 Grace Blackwell Ultra – a 72‑core Grace processor + Blackwell Ultra graphics chip, connected via NVLink C2C at 900 GB/s. Memory • 784 GB total memory. • CPU: 496 GB LPDDR5X (396 GB/s). • GPU: 252 GB HBM3e (7.1 TB/s). Both memories are unified, allowing the processor and graphics accelerator to share resources.

2. Expandability
* PCI‑Express – three PCIe 5.0 x16 slots: one full‑size (16 lanes), two others at 8 lanes each.

This allows discrete GPUs for modeling and ray‑tracing tasks. Supported GPUs:

* RTX Pro 6000 Workstation Edition

* RTX Pro 6000 Blackwell Max‑Q Workstation Edition

* RTX Pro 4000 Blackwell SFF Edition

* RTX Pro 2000 Blackwell

* Additional slots – four M.2, audio jacks, and USB ports.

3. Networking
* Controller – Nvidia ConnectX‑8 SuperNIC with up to 800 Gb/s via two QSFP112 ports.

* Enables linking up to two DGX stations for scaling AI projects.

4. Power
Port Quantity ATX (24‑pin)1 EPS (8‑pin)1 GPU power 12 V‑2x63
Total system power – 1600 W.

5. Availability
* The DGX Station is already available for order and will ship in the coming months through partners: Asus, Dell, Gigabyte, MSI, Supermicro, and HP.

Thus Nvidia offers a compact yet high‑performance workstation capable of meeting the demands of even the most demanding AI applications.