© ROOT-NATION.com - Use of content is permitted with a backlink.
ASUS is pushing the boundaries of what technology can do with the announcement of the ExpertCenter Pro ET900N G3 desktop computer. This device looks like a classic tower-style system unit, but inside it hides the ultra-powerful NVIDIA GB300 Grace Blackwell Ultra desktop-class superchip. The main purpose of this new product is to enable local design of artificial intelligence systems, training of various models, and execution of inference computations. This gives users the performance of a full-fledged data center while completely eliminating the need for a dedicated server room.
Read also: AERONAUT – everything that flies above the ground: aviation, UAVs and drones, rockets, and space
At first glance, the device resembles a perfectly ordinary workstation measuring approximately 23.2×58.4×56.5 cm, which can easily be placed on a standard desk. Despite its understated exterior, the internal components are impressively powerful. The hardware platform combines a Grace central processor, featuring 72 cores based on the Arm architecture, and the advanced Blackwell Ultra graphics chipset, which together deliver up to 20 petaflops of computing power for artificial intelligence tasks.
One of the system’s most striking features is its memory subsystem, with a total capacity of a colossal 748 GB of unified coherent memory. This configuration consists of 496 GB of high-speed LPDDR5X RAM dedicated to the CPU, as well as 252 GB of high-bandwidth HBM3e memory, integrated directly into the graphics adapter. The resulting architecture enables interaction with extremely large AI models, including those approaching a trillion parameters, all directly from the user’s workstation.
To ensure sustained operation at maximum performance without the risk of performance drops due to overheating, the designers equipped the device with an uncompromising cooling system and a robust power supply rated for 1600 W. Options for future upgrades and expansion include three PCIe 5.0 slots, as well as slots for connecting up to four NVMe SSDs in the M.2 2280 form factor. Network connectivity is provided via ultra-high-speed NVLink-C2C and ConnectX-8 interfaces. The computer comes with the Ubuntu operating system pre-installed and a suite of built-in software tools from NVIDIA for AI work, while the developers promise to add support for the Windows family at a later date. The external interface set is quite practical for everyday use: users have access to eight USB 3.2 ports, a single USB 2.0 port, and a pair of Ethernet ports with data transfer speeds of up to 10 Gbps each.
The ExpertCenter Pro ET900N G3 is already available for retail purchase in the United States, with a starting price set at a hefty $9,999. Clearly, this product was designed with researchers, software engineers, and commercial organizations in mind that specialize in generative AI, large language models, complex simulation processes, and autonomous systems. In fact, ASUS has managed to pack the philosophy and capabilities of the well-known NVIDIA DGX Station systems into a significantly more compact form factor.
The emergence of such a device clearly reflects current industry trends aimed at making top-tier hardware for artificial intelligence computing more accessible for desktop use. Certainly, the price is very high, but for teams of specialists who have an urgent need for powerful local computing resources and wish to avoid the complexities associated with deploying and maintaining traditional server infrastructure, this computer has the potential to be a truly game-changing solution.
Read also:
- ASUS Zenbook A16 UX3607 Review: Bringing Qualcomm to Mainstream
- ASUS ROG Strix Morph 96 Wireless Review: Hot-Swap Functionality and Thoughtful Customization