Nvidia (NASDAQ:) unveiled on Monday its latest artificial intelligence (AI) chips and software at its developer’s conference in San Jose, marking a significant step in reinforcing its leadership in the AI sector.
The new AI graphics processors, named Blackwell, with the first chip, the GB200, are set to be released later this year.
These chips promise to deliver a substantial performance boost, offering 20 petaflops of AI performance, a significant leap from the 4 petaflops provided by the current H100 model.
The GB200 is designed to enhance the ability of AI companies to develop larger and more complex models, thanks to its built-in transformer engine, which is optimized for running transformers-based AI, crucial for technologies like ChatGPT.
“Hopper is fantastic, but we need bigger GPUs,” Nvidia CEO Jensen Huang at the company’s GTC conference in California.
The Blackwell GPU, a large chip that combines two dies into one unit manufactured by TSMC, will also be available as part of a complete server setup, the GB200 NVLink 2, which integrates 72 Blackwell GPUs.
Cloud services from major companies like Amazon, Google, Microsoft, and Oracle will offer access to the GB200, facilitating the development and training of AI models at scale.
In particular, Amazon Web Services plans to establish a server cluster featuring 20,000 GB200 chips, enabling the deployment of models with up to 27 trillion parameters—far exceeding the capacity of current models such as GPT-4.
“For three decades we’ve pursued accelerated computing, with the goal of enabling transformative breakthroughs like deep learning and AI,” said Huang
“Generative AI is the defining technology of our time. Blackwell is the engine to power this new industrial revolution. Working with the most dynamic companies in the world, we will realize the promise of AI for every industry,” he added.
The company also announced a new software solution named NIM, aimed at simplifying AI deployment. This revenue-generating tool provides an added incentive for customers to continue using Nvidia chips.
Read the full article here