Asus launches compact supercomputer for AI tasks

Asus has launched a new mini supercomputer called the Ascent GX10. This device is aimed at helping with artificial intelligence (AI) and machine learning tasks. The Ascent GX10 runs on Nvidia’s GB10 Grace Blackwell Superchip. This chip combines a powerful 20-core CPU with a Blackwell GPU. Together, they provide up to 1,000 Tera Operations Per Second (TOPS) of AI performance. This performance makes it ideal for tasks that require high computing power, such as AI model training and edge computing. This mini supercomputer also includes 128GB of unified memory. This allows users to work with large AI models, including those with up to 200 billion parameters. For even larger applications, two units can be linked together. This setup can handle models with up to 405 billion parameters, like Meta's Llama 3.1. Traditionally, high-performance AI computing needs large data centers because of power, space, and cooling requirements. The Ascent GX10 offers a compact solution that lets researchers and developers run complex AI projects right from their desks. Asus aims to meet the needs of AI startups, research centers, and businesses with this flexible system. The GX10 can function as a standalone workstation or fit into private cloud environments. Pricing and availability details for the Ascent GX10 have not been released yet. However, it is expected to be similar in price to Nvidia’s Project Digits, which is around $3,000. Both devices aim to make high-performance AI computing available in smaller spaces. The Ascent GX10 is designed for professionals looking for local AI computing power without needing to use cloud services.