What if NVIDIA’s strongest symbiosis of CPU and GPU design, initially meant for server infrastructure, is put inside a desktop kind issue? That’s precisely the query ASUS requested and determined to do. In its ExpertCenter Professional ET900N G3, ASUS utilized the NVIDIA GB300 Grace Blackwell Extremely desktop superchip, delivering 20 PetaFLOPS of FP4 precision compute, alongside 784 GB of unified cache-coherent reminiscence, 288 GB of that are attributed to HBM3E on the GPU itself, and the remaining 496 GB to LPDDR5X embedded across the Grace CPU. With the GB300 “Blackwell Extremely,” NVIDIA permits OEM companions to make the most of its DGX blueprint and design their very own desktops, so ASUS is enterprise a customized implementation right here. To attach the monstrous compute to the skin world, ASUS embeds ConnectX-8 SuperNIC, which runs at 800 Gb/s, and installs NVIDIA’s DGX OS, a Ubuntu fork with NVIDIA-specific kernel tweaks, optimizations, and extra drivers.
The motherboard powering the system is designed for severe expandability, providing three full-length PCIe x16 slots that permit you to stack a number of GPUs or specialised accelerators, together with additional RTX playing cards for max compute efficiency, and three M.2 growth slots for high-speed SSDs. Energy supply is attention-grabbing because it contains the usual ATX and EPS12V connectors for the CPU and system, plus three devoted 12V‑2×6 GPU energy connectors able to supplying as much as 1,800 Watts. Pricing and availability are unknown, however count on to pay $30,000+ for a system of this caliber, which is particularly aimed toward AI researchers operating these bins as a part of their bigger infrastructure.
