Public backlash against data centers has emphasized their water and energy consumption, and now Nvidia is highlighting its claim that the Rubin generation reference design for a fully liquid-cooled data center has “eliminated massive amounts of power use and almost all water use.” Still, this does not address all of the concerns surrounding AI data centers, including the costs incurred during their construction and the power generation requirements of the huge facilities. Also in the form of gizmodo Pointingly, Nvidia’s blog post does not mention the cost of building this style of data center versus one that uses less efficient air cooling, but claims that “each cloud provider and data center operator is required to build [Rubin] Is making change.”
The increase in efficiency is partly due to running the AI servers warmer, up to 113 °F (45 °C). In a recent report, Amazon claimed higher heat tolerance as part of making its mostly air-cooled data centers more efficient.
With Nvidia’s system, “heat is captured directly on the chip and transported through a liquid loop operating at very high temperatures, allowing external dry coolers to efficiently reject heat for most of the year,” with a lot more flexibility when it comes to ambient air temperatures.
According to Josh Parker, Nvidia’s sustainability lead, the reference design takes water use “from about 2.6 million gallons per MWh per year for traditional cooling-tower-based systems to near zero – a 100 percent reduction.”
<a href