Nvidia’s H100 GPUs will consume more power than some countries — each GPU consumes 700W of power, 3.5 million are expected to be sold in the coming year

Key Points:

  • Nvidia’s H100 AI GPUs are set to consume a staggering amount of power, with projections suggesting their aggregate consumption could rival major American cities and even surpass that of small European countries.
  • Paul Churnock of Microsoft forecasts that the power consumption of Nvidia’s H100 GPUs, when deployed en masse by late 2024, will exceed that of all households in Phoenix, Arizona, making it the fifth-largest electricity consumer in the U.S.
  • Despite the jaw-dropping power consumption, ongoing advancements in AI and HPC GPU efficiency offer a silver lining, with upcoming GPUs like Nvidia’s Blackwell-based B100 promising higher performance and more efficient energy usage.

Summary:

In the glitzy world of tech, Nvidia is making headlines not just for its cutting-edge H100 AI GPUs but also for the mind-boggling power they consume. These GPUs, which boast a peak power consumption of 700W, are forecasted to collectively guzzle up more electricity than the average American household, with estimates suggesting their impact will rival that of a major American city or even surpass the combined consumption of small European countries! Paul Churnock, a luminary at Microsoft, predicts that by late 2024, when millions of these GPUs will be in operation, their combined consumption will surpass that of all households in Phoenix, Arizona. By then, these power-hungry contraptions might just clock in as the fifth largest electricity consumer in the U.S., nimbly squeezing between the energy appetites of Houston and Phoenix.

 

However, amidst the power consumption histrionics, there’s an important kernel of optimism. The relentless march of progress in AI and HPC GPU efficiency is poised to temper these staggering energy demands. Even as Nvidia’s upcoming Blackwell-based B100 threatens to outpace the voracious power appetite of the H100, it promises to offer superior performance, resulting in more work done for each unit of power consumed.

DAILY LINKS TO YOUR INBOX

PROMPT ENGINEERING

Prompt Engineering Guides

ShareGPT

 

©2024 The Horizon