From NVIDIA:

Generative AI is set to revolutionize various industries with NVIDIA introducing the Blackwell computing platform. CEO Jensen Huang highlighted the platform during the keynote address at the GTC conference, emphasizing accelerated computing’s potential in driving down costs and improving performance across sectors. The event showcased major advancements in AI technology, including real-time generative AI and advanced simulation capabilities, paving the way for trillion-parameter large language models.

Huang unveiled the NVIDIA Blackwell platform designed to enhance AI infrastructure by delivering real-time generative AI capabilities. The platform, named after mathematician David Blackwell, boasts significant performance improvements in both training and inference tasks. With a focus on scaling up computing power, NVIDIA introduced the GB200 Grace Blackwell Superchip for increased efficiency and advanced networking speeds.

Industry leaders praised Blackwell, with endorsements from Alphabet, Amazon, Google, Microsoft, and others. Huang emphasized the transformative impact of generative AI on software development, shifting towards assembling AI models rather than traditional coding methods. The NVIDIA NIMs microservices offer a new way to create software, leveraging accelerated computing libraries and generative AI models to streamline application development processes.

NVIDIA also introduced the DGX SuperPOD powered by the GB200 Grace Blackwell Superchips, designed for processing trillion-parameter models in AI supercomputing applications. The liquid-cooled rack-scale architecture offers exaflops of AI performance and fast memory capabilities, catering to the increasing demand for AI-driven solutions in various industries. The AI revolution sparked by generative AI and accelerated computing heralds a new era of innovation and efficiency in technology.



Read more at NVIDIA: ‘We Created a Processor for the Generative AI Era,’ NVIDIA CEO Says