Nvidia unveils powerful new chips to support the AI sector

19 Mar 2024

Image: © BINGJHEN/Stock.adobe.com

Nvidia is releasing the B200 GPU and the GB200 ‘superchip’ later this year, as it focuses on retaining its dominance in the AI chips market.

Nvidia has revealed new GPU architecture called Blackwell that it claims includes the most powerful AI chip in the world.

The company said its Blackwell architecture GPUs will be powerful enough to enable AI training and real-time large language model inference for models containing up to 10trn parameters.  Nvidia also said companies will be able to run models with massive reductions in cost and energy consumption thanks to its new designs.

Meanwhile, the company also shared details of its new “superchip” – the GB200 Grace Blackwell – which connects two NVIDIA B200 Tensor Core GPUs to a NVIDIA Grace CPU.

Using the GB200, the company said it will offer rack-scale systems that combine dozens of these superchips to manage the “most compute-intensive workloads”. Nvidia claims this system will reduce cost and energy consumption by 25 times.

“For three decades we’ve pursued accelerated computing, with the goal of enabling transformative breakthroughs like deep learning and AI,” said Nvidia founder and CEO Jensen 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 realise the promise of AI for every industry.”

The Blackwell-based products will be available later this year, with various cloud giants will be among its customers – including Amazon Web Services, Google Cloud, Microsoft Azure and Oracle.

Nvidia appears to be pushing these Blackwell products as the successors to its powerful AI chips, which many AI companies have used to support the development of their models.

The company’s focus on the AI chips market helped its profits soar in 2023, as the company reached a $1trn valuation at the end of May 2023. Nvidia hit the $2trn valuation mark for the first time last month.

But the company is also facing its own share of legal troubles. Earlier this month, a trio of authors sued Nvidia, claiming the company used copyrighted books to train one of its AI models.

Find out how emerging tech trends are transforming tomorrow with our new podcast, Future Human: The Series. Listen now on Spotify, on Apple or wherever you get your podcasts.

Leigh Mc Gowran is a journalist with Silicon Republic

editorial@siliconrepublic.com