In a world of rapidly advancing AI technology reliant on massive data centres, NVIDIA Corp (NASDAQ:NVDA) has taken AI infrastructure to a new level with the launch of the fourth-generation DGX H100 System.
These systems provide 32 petaflops of AI performance potential, capable of performing one quadrillion (1015) floating-point operations (flops) per second.
In data storage terms, that’s the equivalent of just under 212,766 DVDs.
The new system performs at a rate six times greater than its predecessor, capable of handling the enormous burden of large language models needed for generative AI applications.
The company intends to combine a total of 576 DGX H100 units using its DGX SuperPOD technology into a supercomputer – Eos.
Ranked #9 in the TOP500 list of fastest supercomputers, Eos is the culmination of our ongoing commitment to pushing the boundaries of #AI technology and infrastructure. Learn more: https://t.co/gJViuPY0Ip #DataCenter #NVIDIADGX pic.twitter.com/AUBfNU7GmR— NVIDIA (@nvidia) February 15, 2024
Major step up in computing power
“AI has fundamentally changed what software can do and how it is produced. Companies revolutionising their industries with AI realize the importance of their AI infrastructure,” said Jensen Huang, founder and CEO of NVIDIA.
“Our new DGX H100 systems will power enterprise AI factories to refine data into our most valuable resource — intelligence.”
Nvidia expects its Eos supercomputer to offer 18.4 exaflops of data, 18,400 petaflops of computing power, placing it as the fastest AI system in the world once it begins operation later this year.
The closest supercomputer currently operating is the Fugaku supercomputer in Japan, which is capable of only a quarter of that performance.
The company says Eos will serve as a blueprint for advanced AI infrastructure from NVIDIA, as well as its OEM and cloud partners.