ASUS Introduces Ascent GX10 AI Supercomputer Powered by NVIDIA GB10 Grace Blackwell Superchip
Petaflop-scale AI computing in a compact package
- Incredible AI performance: Up to 1,000 AI TOPS processing and 128GB unified memory empower seamless model development and inferencing
- Cutting-edge architecture: NVIDIA® GB10 Grace Blackwell Superchip with a Blackwell GPU, 20-core Arm CPU and NVLink®-C2C boosts AI workloads
- Powerful, scalable and local: Configure as standalone or private-cloud system to drive AI models on desktop, or connect two GX10s via ConnectX-7 NICs
TAIPEI, Taiwan, March 18, 2025 — ASUS today announces its groundbreaking AI supercomputer, ASUS Ascent GX10, powered by the state-of-the-art NVIDIA® GB10 Grace Blackwell Superchip. This revolutionary device places the formidable capabilities of a petaflop-scale AI supercomputer directly onto the desks of developers, AI researchers and data scientists around the globe.
As the size and complexity of generative AI models grow, local development efforts face increasing challenges. Prototyping, tuning and inferencing large models require substantial memory and compute performance. To address these needs, Ascent GX10 is designed to provide developers with a powerful, economical desktop solution for AI development.
Unparalleled AI performance
The forthcoming ASUS Ascent GX10 delivers up to 1,000 TOPS[i] of AI performance to power large AI workloads. With 128GB of coherent unified system memory, developers can experiment, fine-tune or inference the latest generation of reasoning AI models of up to 200-billion parameters.
Powered by NVIDIA Grace Blackwell technology
Central to the new ASUS Ascent GX10 is the advanced NVIDIA GB10 Superchip, designed with the Grace Blackwell architecture and optimized for compact-size form factor. The enormously-powerful GB10 chip includes a robust Blackwell GPU with fifth-generation Tensor Cores and FP4 support, delivering up to 1000 TOPS of AI processing power. Additionally, it features a high-performance Grace 20-core Arm CPU that enhances data preprocessing and orchestration, accelerating the tuning of models and real-time inferencing. The GB10 Superchip employs NVIDIA NVLink®-C2C to provide a cohesive CPU+GPU memory model with five times the bandwidth of PCIe® 5.0.
Handling cutting-edge large-parameter generative-AI models
ASUS Ascent GX10, with its 128GB of unified system memory and FP4 data format support, is capable of managing AI models up to 200-billion parameters. This allows AI developers to prototype, fine-tine and inference the latest AI reasoning models — up to 70-billion parameters — directly on their desktop. Thanks to the integrated NVIDIA ConnectX® network interface cards (NICs), two GX10 systems can be linked to handle even larger models, such as Llama 3.1 with 405-billion parameters.
Local development, scalable deployment
ASUS Ascent GX10 provides a powerful, economical experimentation platform for developers to prototype models and AI applications, freeing up essential compute resources in their clusters, which are more suited for training and deploying production models. Using NVIDIA AI software enables Ascent GX10 users to seamlessly transition their models from desktop environments to NVIDIA DGX Cloud or any accelerated cloud or data center infrastructure with virtually no code adjustments, facilitating easier prototyping, fine-tuning and iteration processes.
Empowering the AI community
“AI is transforming every industry, and the ASUS Ascent GX10 is designed to bring this transformative power to every developer’s fingertips,” said KuoWei Chao, General Manager of ASUS IoT and NUC Business Group. “By integrating the NVIDIA Grace Blackwell Superchip, we are providing a powerful yet compact tool that enables developers, data scientists, and AI researchers to innovate and push the boundaries of AI right from their desks.”
