Global technology leader ASUS has announced the worldwide availability of the ExpertCenter Pro ET900N G3, an advanced deskside AI supercomputer. Engineered to shift data-center-grade performance directly to local professional settings, the system provides enterprises, developers, and researchers with a powerful solution for high-level model training and execution. Built on the NVIDIA DGX Station GB300 architecture and powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, the platform bridges the gap between massive computational needs and localized on-premises infrastructure.
Localized Computational Power Tailored for Modern Workloads
With increasing adoption of intricate workflows of artificial intelligence in contemporary companies, the use of cloud architecture alone is faced with problems associated with issues of data latency, unanticipated operational costs, and governance risks. The ExpertCenter Pro ET900N G3 addresses this emerging market need through giving companies more control of their data and processing workloads locally.
Key architectural components fueling this local AI framework include:
- AI Multi-Petaflop Performance: This architecture has the capacity to provide up to 20 PFLOPS of computational power dedicated to AI, thus allowing the users to speed up LLM fine-tuning, physical AI engineering, generation workflows, and deep learning research safely in the office setting.
- Large Coherent Memory: With 748GB of large coherent unified CPU-GPU memory powered by NVIDIA NVLink-C2C high bandwidth interconnect technology, the architecture helps to avoid usual workflow latency bottlenecks. It allows developers to perform training and processing of cutting-edge AI models with up to 1 trillion parameters.
- Data Center Capacity for Deskside Deployments: The architecture is specifically designed for offices and incorporates large data center capacity with extremely compact deskside form factor without the necessity of having data centers or costly secondary cooling infrastructure.
Also Read: HPE Expands Self-Driving Networks Across Edge, Campus, Data Center, and AI Factories
Accelerating Next-Generation Autonomous Agents
The system is optimized to accelerate agentic AI paradigms and autonomous workspace assistants. Local benchmarking conducted by ASUS engineering specialists utilizing vLLM showcased the system’s operational speed. During stressful open-source model evaluations using the large-scale Qwen model, the hardware successfully sustained an output throughput of roughly 864 tokens per second, with overall system processing peaking near 1,600 input and output tokens per second combined.
The ExpertCenter Pro ET900N G3 is designed to facilitate special-purpose use cases such as NVIDIA NemoClaw. With its framework, companies can efficiently design, test, and deploy their on-demand intelligent agents in an enterprise setting. Being integrated in totality with the NVIDIA AI software suite, the ecosystem enables smooth development without compromising on data security or deployment options.
