- Asus Expercener Pro et900n G3 hides the craziest chip in Nvidia
- With up to 784 GB of memory, it manages the models that your RTX 5090 simply cannot
- No rack, no noise – just AI performance of the supercomputer class on a desktop that you can use
Asus has unveiled a new high -performance office PC which offers AI performance on a Petaflop scale, not in a flashy server rack, but in a surprisingly modest chassis.
The Asus Expercener Pro ET900N G3 looks like any standard activity tower – so much so that it even includes a DVD disc and a curious slit that looks like a return in the early 2000s.
At the heart of this professional PC is the NVIDIA GB300 Ultra, a two-part module that combines a Grace CPU with a Blackwell GPU via Nink-C2C, the NVIDIA wide-band interconnection, which makes it ideal for the programming and development of AI.
Designed for serious AI work
The unified chip architecture allows the CPU and the GPU to share a single memory pool, reduce latency and improve the efficiency of Large -scale AI workloads. The system can deliver up to 20 performance pflops for the formation of large languages models or execute inference on high parameter models.
It supports up to 784 GB of coherent memory, more than double the combined VRAM of a workstation with four RTX 6000 ADA cards.
This access scale for memory is essential for developers and researchers working with models that go beyond the capacities of traditional GPUs such as the GeForce RTX 5090, which offers 32 GB of VRAM.
The Expercener Pro ET900N G3 also includes the support for NVIDIA Connectx-8 Supernic, allowing a high speed network between systems. This allows him to operate in clusters or in a larger corporate AI deployment.
Despite its performance, it retains an office shape factor, eliminating the need for rack installation, personalized cooling solutions or infrastructure requests for a data center.
On the software side, the system executes Nvidia DGX OS, a specialized Linux distribution based on Ubuntu adapted to AI workloads. It provides native management for the complete NVIDIA software battery, notably Cuda, Tensorrt and libraries for automatic learning and data science.
It also supports remote scope, allowing the ET900N G3 to integrate transparently with other DGX systems if an additional calculation power is necessary.
