NVIDIA Announces Liquid-Cooled PCIe GPUs For Mainstream Servers

Nvidia has unveiled a new strategy for data centers crunching enormous quantities of data or training AI models that uses less energy: liquid-cooled GPUs. In an effort to stop climate change, Nvidia recently announced two liquid cooled PCIe cards for the A100 and H100 GPU servers. Both the A100 and H100 PCIe cards deliver high performance and energy efficiency while reducing environmental impacts. 

According to Nvidia officials, datacenters consume 1% of the world’s electricity. These new offerings are essential to the market since it has been confirmed that GPUs are more power efficient than CPUs. 

“This marks the first liquid-cooled GPU introduced to our lab, and that’s exciting for us because our customers are hungry for sustainable ways to harness AI,” said Zac Smith, head of edge infrastructure at Equinix. “We’ll turn a waste into an asset.”

After running countless tests, Nvidia determined that a liquid-cooled data center could handle the same workloads as an air-cooled facility while using 30% less energy. It’s estimated that a liquid-cooled data center could hit 1.15 PUE (power usage effectiveness), far below 1.6 for its air-cooled counterpart.

Although the A100 SXM form factor card is already available for liquid-cooled servers, this new PCIe addition changes the game. The upcoming PCIe cards are expected to use direct-to-chip liquid cooling, taking up only one PCIe slot in a server as opposed to two slots with the air-cooled versions. This will allow liquid-cooled GPU servers to be more readily available with over a dozen manufacturers like: ASUS, GIGABYTE, Supermicro, and more helping to implement this shift.

Smith added that today’s liquid-cooled GPUs deliver the same performance for less energy. In the future, they’re expected to provide an option to extract more performance for the same energy expended. 

“Measuring wattage alone is not relevant, the performance you get for the carbon impact you have is what we need to drive toward,” said Smith.

The new cards are just the beginning of what Nvidia has in store for us. Officials say they already have more liquid-cooled server cards on the drawing board and are poised to introduce other targeted components like in-car systems that need to keep cool in enclosed spaces. 

The liquid-cooled A100 accelerator, based on the Ampere architecture, will be available to customers in Q3 of this year. Meanwhile, the H100 accelerator, based on the Hopper architecture, will begin to ship in early 2023. 

Need Nvidia GPUs now? Check out our website! Don’t see what you need? Contact us today! 1-800-237-0402.