TLDR: ASUS has launched the ExpertCenter Pro ET900N E3, a desktop PC featuring NVIDIA’s GB300 Blackwell superchip, capable of delivering up to 20 PFLOPS of AI processing power and 784GB of unified coherent memory. This system aims to bring data center-level AI capabilities to a desktop form factor, enabling complex AI training and scientific simulations locally.
ASUS has introduced a groundbreaking addition to its ExpertCenter lineup with the ExpertCenter Pro ET900N E3, a desktop PC that redefines the boundaries of on-premises artificial intelligence development. This innovative system integrates NVIDIA’s revolutionary GB300 Blackwell superchip, a technology typically reserved for large-scale data centers, into a compact desktop chassis. The GB300 superchip combines a Grace CPU and a B300 GPU, delivering an astonishing 20 petaFLOPS of AI processing performance and 784GB of unified coherent memory. This immense power is sufficient to train complex large language models (LLMs) and execute real-time scientific simulations that traditionally necessitate extensive cloud infrastructure.
The ExpertCenter Pro ET900N E3 is engineered with enterprise-grade thermal solutions and an industrial design to ensure sustained performance without the typical acoustics associated with server rooms. The system also includes NVIDIA’s DGX OS, a comprehensive AI software suite, transforming the desktop into a turnkey development hub for AI researchers and engineers. Industry analysts view this as a significant shift towards ‘deskside supercomputing,’ allowing professionals to bypass cloud latency for sensitive projects.
Delving into the specifications, the ET900N E3 boasts a sophisticated memory architecture: 496GB of LPDDR5X RAM for the Grace CPU synergizes with 288GB of HBM3E memory for the B300 GPU, facilitating seamless data sharing. For connectivity, it features NVIDIA ConnectX-8 SuperNIC, providing an impressive 800Gb/s bandwidth for multi-node cluster support. The system offers robust expandability with three PCIe Gen5 x16 slots and triple M.2 NVMe slots, accommodating additional GPUs or high-speed storage. Power delivery is managed by three industrial-grade 16-pin connectors, supplying up to 1,800 watts, a capacity comparable to small server racks.
Also Read:
- Huawei Unveils CloudMatrix 384 AI Computing System, Challenging Nvidia’s Dominance
- NVIDIA Unveils Llama Nemotron Super v1.5: Advancing AI Reasoning and Agentic Capabilities
NVIDIA’s decision to allow OEM partners like ASUS to utilize its DGX blueprint for designing their own desktops marks a strategic shift, with GB300 servers expected to enter mass production by Q4 2025. While pricing and availability for the ExpertCenter Pro ET900N E3 are yet to be officially announced, industry experts anticipate a price point upwards of $30,000, targeting AI researchers and developers who require such high-caliber systems as part of their larger infrastructure.
