Nvidia unveils latest supercomputer for generative AI

29 May 2023

Image: © MichaelVi/Stock.adobe.com

The latest Nvidia DGX GH200 supercomputer has nearly 500 times more memory than the previous model unveiled in 2020.

Nvidia CEO Jensen Huang made a spate of company announcements today (29 May), including details about a new supercomputer that will help companies develop generative AI models.

Huang said that the new Nvidia DGX GH200 supercomputer connects 256 of its Grace Hopper superchips into a single massive 144TB graphics processing unit (GPU) that can deliver up to 1 exaflop of computing power for giant generative AI models, among other applications.

This system has nearly 500 times more memory than the previous Nvidia DGX supercomputer introduced in 2020.

“Generative AI, large language models and recommender systems are the digital engines of the modern economy,” Huang said in a keynote at the company’s Computex conference in Taiwan.

“DGX GH200 AI supercomputers integrate Nvidia’s most advanced accelerated computing and networking technologies to expand the frontier of AI.”

Google Cloud, Meta and Microsoft are among the first companies expected to gain access to the latest supercomputer to explore its capabilities for generative AI workloads as competition in the space continues to stiffen.

News of the latest supercomputer comes as the Santa Clara-based company exceeded expectations in its latest earnings report last week, with a forecast to earn $11bn in revenue for its next fiscal quarter due to “surging demand” for its AI and data centre products.

Huang said during the earnings call that the computer industry is going through two “simultaneous transitions”, which are accelerated computing and generative AI.

“A trillion dollars of installed global data centre infrastructure will transition from general purpose to accelerated computing as companies race to apply generative AI into every product, service and business process,” Huang said.

Earlier this month, Nvidia revealed that Google Cloud is using its L4 GPU and Vertex AI to help companies develop generative AI applications. Nvidia is also in a multi-year collaboration with Microsoft in a bid to build one of the world’s most powerful AI supercomputers.

The company is also building Israel’s most powerful AI supercomputer called Israel-1 to meet soaring demand for generative AI applications, with companies based in the country getting first access to the system that, according to Reuters, cost hundreds of millions of dollars.

10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.

Vish Gain is a journalist with Silicon Republic