Nvidia says its AI data center design is faster to use less water

The public push against data centers has emphasized theirs water and energy useand now Nvidia is to highlight His claim that Rubin’s design of the water-cooled storage facility “eliminates high energy and water consumption.” However, it does not address all the challenges surrounding AI data centers, including their construction, and for required to produce energy of large equipment. Also, if Gizmodo points out, Nvidia’s blog post does not address the cost of building this style of data center vs. one using air conditioning, but says “every cloud provider and data center provider to (Rubin) is changing.”

Part of the gain is due to running the AI ​​servers at higher temperatures, up to 113 degrees Fahrenheit (45 degrees Celsius). In a recent report, Amazon similarly to be accepted high temperature tolerance as part of its highly ventilated data center design.

With Nvidia’s system, “heat is taken directly from the chip and transferred through liquid loops that operate at high temperatures, allowing external coolers to resist heat well for years,” and is highly variable when it comes to the temperature of the surrounding air.

According to Nvidia’s director of security, Josh Parker, the design uses water “from about 2.6 million gallons per megawatt per year to run the cooling towers to zero – to a 100 percent reduction.”