Tech News
← Home  ·  All topics

Nvidia Dgx Station

2 GoKawiil briefs on this topic

NVIDIA demos GLM-5.3 served at 607.8 tok/s across two DGX Stations with vLLM

NVIDIA published a tutorial showing how to serve Z.ai's GLM-5.3 NVFP4 model, a 753-billion-parameter mixture-of-experts model with 40 billion parameters activated per token, across two GB300 DGX Stations linked via ConnectX-8 RDMA. Because the roughly 464 GB quantized checkpoint exceeds the 252 GB of HBM3e on a single GB300 GPU, the setup uses tensor parallelism (TP=2) and vLLM to split weights and computation across both machines while exposing one OpenAI-compatible API endpoint, reportedly measuring throughput of 607.8 tokens per second.

Nvidia's GB300 DGX Station desktop appears listed at $94,930

An Exxact listing has confirmed pricing for Nvidia's DGX Station desktop tower, the Valence VWS-158270643, at $94,930 to start. The system pairs a GB300 Superchip—Blackwell Ultra GPU with 252GB of HBM3e memory—with a 72-core Grace CPU over NVLink-C2C, adding 496GB of LPDDR5x RAM for a combined 748GB of unified memory. Nvidia had unveiled the machine at Computex 2026 without disclosing a price, leaving partners like Dell, HP and Exxact to quote figures individually.