- The NVIDIA DGX station is powered by the GB300 Grace Blackwell Ultra
- The OEM make their own versions – Dell’s is Pro Max with GB300
- The next HP GB300 workstation will be the G1N station of ZGX Fury Ai G1N
NVIDIA has unveiled two superordinators from the DGX Person fed by its Grace Blackwell platform.
The first is DGX Spark (previously called Project Digits), an AI compact supercomputer that works on the GB10 Grace Blackwell Superchip from Nvidia.
The second is DGX Station, a SuperComputer class workstation that looks like a traditional tower and is built with the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip.
Dell and HP reveal their versions
The GB300 offers the latest generation tensor nuclei and the FP4 accuracy, and the DGX station includes 784 GB of coherent memory space for large-scale training and inference workloads, connected to a Grace CPU via NVLink-C2C.
The DGX station also includes the Supernic Connectx-8, designed to overeat hyperscal AI IT workloads.
NVIDIA – Asus, HP and DELL’s OEM partners produce DGX Spark rivals powered by the same GB10 Superchip. HP and Dell also prepare competitors at DGX station using the GB300.
Dell has shared new details on its next AI work station, the Pro Max with GB300 (its DGX Spark version is called Pro Max with GB10).
The specifications of his supercomputer’s class workstation include 784 GB of unified memory, up to 288 GB of HBM3E GPU memory and 496 GB of LPDDR5X memory for the CPU.
The system offers up to 20,000 summits of FP4 calculation performance, which makes it well suited to LLMS training and inference with hundreds of billions of parameters.
The HP version of the DGX station is called the ZGX Fury AI G1N station. Z by HP is now one of the company’s product ranges, and the “N” at the end of the name means that it is powered by an NVIDIA processor – in this case, the GB300.
HP says that the AI station of ZGX Fury G1n “provides everything necessary for the AI teams to build, optimize and evolve models while retaining safety and flexibility”, noting that it is integrated into the ecosystem of the AI wider HP station, next to the ZGX Nano Nano alternative (its alternative DGX Spark).
HP also extends its software tools and its AI support offers, providing resources designed to rationalize the productivity of the workflow and improve the development of local models.
The price of the DGX station and the DELL and HP workstations is not yet known, but they will obviously not be cheap. The price of the tiny DGX sparkle starts at $ 3,999, and the biggest machines will cost much more.
