ASUS Ascent GX10, GB10 Grace Blackwell Desktop AI System

A compact desktop AI development system, a small, power-efficient personal computer for AI development, not a rack server despite the "AI Supercomputer" name. It is built on the NVIDIA GB10 Grace Blackwell Superchip with 128GB of unified memory and 1 petaFLOP of AI performance, and it sits on a desk. It is ASUS's version of NVIDIA's GB10 desktop platform, for fine-tuning and running large AI models locally.

They share the same core. NVIDIA licensed partners to build systems around its GB10 Grace Blackwell Superchip, and both the ASUS Ascent GX10 and NVIDIA's own DGX Spark are built on that same chip with 128GB of memory. The GX10 is ASUS's implementation, distinguished by ASUS's build and thermal design and its dual-unit stacking. For a buyer choosing between GB10 desktop systems, they are close alternatives, and MillionMiner can weigh the GX10 against the DGX Spark for your needs.

No. Despite the "supercomputer" name, the GX10 is a personal desktop development machine, not a rack-mounted server or data center system. The data center systems (the H100, H200, B200 hardware) draw kilowatts in a rack and train at scale; the GX10 sits on a desk, draws a fraction of the power, and is for local development. It is the desk-side counterpart to those systems, not one of them.

It lets two GX10 units work together. Using NVIDIA's NVLink-C2C and ConnectX-7 networking, two GX10s can be linked to combine their resources, taking on larger models than a single unit handles. So a developer can start with one GX10 and add a second to scale local capability. This stacking is one of the GX10's distinguishing features as an ASUS-built GB10 system.

Large AI models, locally. With 128GB of unified memory and 1 petaFLOP of AI performance, it can fine-tune and run inference on large models, with support for up to 200 billion parameters on the device (more across two stacked units). That is well beyond what a normal workstation handles, which makes it a serious local AI development tool for researchers and developers.

Through a shared software stack and workflow. The GX10 runs the NVIDIA AI software stack, so a developer fine-tunes and tests locally on the GX10, then hands the work off to a data center or cloud system for full-scale training or production deployment, without re-engineering the software. Build on the GX10, scale on the data center systems.

The NVIDIA GB10 Grace Blackwell Superchip, 128GB of unified memory, 1 petaFLOP of AI performance, support for models up to 200 billion parameters, NVIDIA NVLink-C2C, ConnectX-7 networking, dual-unit stacking, and ASUS's thermal design, in a compact desktop form. MillionMiner confirms the full configuration, including storage, ports, and exact power, in the quote.

AI researchers, developers, and labs that want serious local AI capability, to fine-tune, experiment with, and run inference on large models at the desk, rather than queuing for shared data center resources or sending everything to the cloud. Its compact footprint and thermal design suit labs, offices, and edge locations, and the stacking option lets a team scale local capability with a second unit.

Yes. As a current product, the ASUS Ascent GX10 ships new. MillionMiner confirms the configuration in the quote and tests every unit before shipment.

Submit your needs through the quote form. A MillionMiner specialist confirms the configuration and quantity, with delivery, and can weigh the GX10 against NVIDIA's DGX Spark and other GB10 systems, or the data center hardware the GX10 hands off to if you also need full-scale training. Every unit is tested before shipment and shipped worldwide DDP with duties handled.