Nvidia and other computer manufacturers have introduced new systems powered by the Nvidia Blackwell architecture at Computex. These systems, featuring Grace CPUs and Nvidia networking, are designed to support (artificial intelligence) AI factories and data centers for generative AI (GenAI) advancements.
“From server, networking, and infrastructure manufacturers to software developers, the whole industry is gearing up for Blackwell to accelerate AI-powered innovation for every field,” said Jensen Huang, founder and CEO of Nvidia.
During his keynote, Huang also announced partnerships with leading manufacturers, including ASRock Rack, Asus, Gigabyte, and Supermicro, to deliver AI systems for cloud, on-premises, embedded, and edge applications. Huang emphasized the shift from traditional data centers to AI factories, highlighting the industry’s readiness to embrace Blackwell technology to accelerate AI innovation.
The new offerings will cater to various applications, ranging from single to multi-GPU setups, x86 to Grace-based processors, and different cooling technologies. The Nvidia MGX modular reference design platform now supports Blackwell products, including the new Nvidia GB200 NVL2 platform. This platform is optimized for large language model inference, data processing, and retrieval-augmented generation, providing significant performance and energy efficiency improvements.
Support from AMD and Intel for the MGX architecture, with their next-generation CPU designs, further enhances the ecosystem. The GB200 NVL2 platform integrates seamlessly into existing data center infrastructures, supporting a range of configurations and networking options.
Nvidia’s partner ecosystem includes TSMC and other key component providers, ensuring rapid development and deployment of new data center infrastructure. Enterprises can leverage Nvidia’s AI Enterprise software platform for creating and running production-grade generative AI applications, supported by Nvidia’s comprehensive technology suite, including Quantum-2 and Spectrum-X networking.