ASUS Ascent GX10 powers local AI image generation

A test of the latest AI image generators was carried out on an ASUS Ascent GX10 workstation equipped with an NVIDIA GB10 SoC that combines a 20‑core Grace CPU and a Blackwell GPU with 128 GB of unified memory. The system successfully ran large models such as FLUX.2‑dev, which requires 64.4 GB of memory, without the subscription fees or usage limits of commercial cloud services, offering full privacy and unlimited generation.

The evaluation highlighted major hardware advantages but also noted significant installation hurdles. ARM64‑compatible Docker containers are scarce, and some dependencies (e.g., torchaudio) required work‑arounds. The team used a native Python environment and an auxiliary AI assistant, Claude Code, to resolve compatibility errors. For user interfaces, the node‑based ComfyUI proved flexible but complex, while the simpler web‑based SwarmUI offered a more accessible workflow.

The findings suggest that powerful, unified‑memory platforms can replace cloud‑based image generation for users who can afford the hardware, though the setup process remains a barrier for most practitioners.