NVIDIA and top computer manufacturers at Computex 2024 introduced systems powered by NVIDIA’s Blackwell architecture, featuring Grace CPUs and advanced networking.

These systems aim to transform data centers into AI factories, driving the next wave of generative AI innovations.

NVIDIA CEO Jensen Huang highlighted the industry’s shift towards accelerated computing and AI factories, emphasizing the role of NVIDIA’s technology in this transformation. The new systems will cater to various applications with configurations ranging from single to multi-GPU setups, x86 to Grace-based processors, and air to liquid cooling.

To support diverse needs, the NVIDIA MGX™ modular reference design now includes Blackwell products, such as the GB200 NVL2 platform, which excels in data analytics and large language model inference. This platform boosts data processing speed by up to 18 times with 8 times better energy efficiency compared to traditional x86 CPUs.

NVIDIA’s MGX architecture allows manufacturers to create over 100 system configurations efficiently, cutting development costs by up to 75% and reducing development time by two-thirds. Major CPU manufacturers AMD and Intel are also contributing to MGX with their own processor modules.

The Blackwell ecosystem includes partners like TSMC and global electronics makers, ensuring rapid development and deployment of new AI data center infrastructure. Enterprises can leverage NVIDIA AI Enterprise software, including NIM™ inference microservices, to build and run generative AI applications.

Taiwanese companies are early adopters, with Chang Gung Memorial Hospital and Foxconn integrating Blackwell for biomedical research and AI-powered smart solutions, respectively, enhancing patient care and personalized customer experiences.

Leave a Reply

Your email address will not be published