en.Wedoany.com Reported - ASUS has launched the ASUS ExpertCenter Pro ET900N G3, a desktop AI supercomputer powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, built on the NVIDIA DGX Station GB300 architecture. It is designed to bring data center-level AI performance to the desktops of enterprises, AI developers, researchers, and data scientists.
As AI rapidly proliferates across industries, organizations increasingly demand local AI infrastructure to support large-scale training, inference, and autonomous agent AI workflows, while reducing reliance on cloud computing. The ET900N G3, through local deployment, enables enterprises to meet AI capability deployment needs with greater control, lower latency, predictable operational costs, and enhanced data privacy. The system is now available globally.
The ET900N G3 supports workloads such as large language model fine-tuning, generative AI, physical AI, deep learning research, and autonomous AI agents. Its GB300 superchip is connected via NVIDIA NVLink-C2C high-bandwidth interconnect technology, enabling high-speed coherent memory access. The system is equipped with 748GB of coherent unified memory, allowing developers to process larger AI models locally. This model achieves AI performance of 20 PFLOPS, meeting demanding development and inference tasks.
Based on the NVIDIA DGX Station GB300 architecture, the ET900N G3 integrates supercomputer-level AI performance into a desktop design, enabling AI infrastructure deployment without the need for a dedicated data center environment. The system is optimized for local AI development, supports the NVIDIA AI software stack, provides a turnkey environment for machine learning, analysis, and experimentation, and supports interconnected AI workflows to scale computing performance.
To support the local execution of agent AI and autonomous AI assistants, the ET900N G3 enables enterprises to maintain strict governance over sensitive data, accelerate real-time operations, and reduce dependence on cloud infrastructure. Stress tests conducted by the ASUS engineering team using vLLM showed that when running the large-scale Qwen open-source AI model, the system achieved an output throughput of approximately 864 tokens per second, with a total input and output processing rate of approximately 1,600 tokens per second. The ET900N G3 also supports the NVIDIA NemoClaw workflow for building and deploying always-on AI assistants and agents in local environments. Combined with the NVIDIA AI software stack, it simplifies development while maintaining enterprise-level control and privacy.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
