Nvidia Develops Innovative Cooling Method for AI Data Centres Aiming for Zero Water Usage

The CSR Journal Magazine

Nvidia has announced a new cooling technology for artificial intelligence data centres, claiming it can significantly reduce water consumption to nearly zero. This initiative comes amid increasing energy demands placed on the industry by advanced AI systems. The latest infrastructure, named the Rubin-generation, utilises a fully liquid-cooled design which aims to eliminate conventional air cooling methods and greatly decrease dependence on water-based cooling solutions.

The system operates by directly removing heat at the chip level via liquid circulating through sealed cold plates. The coolant captures heat and continuously circulates in a closed loop, allowing for its repeated use without being depleted. Nvidia asserts that this method facilitates a departure from traditional cooling towers and systems reliant on fans, which typically lead to higher energy consumption and water use.

Higher Operating Temperatures Enhance Efficiency

A pivotal aspect of Nvidia’s new cooling technology lies in the temperature range at which it operates. The coolant can safely function at temperatures reaching up to 45 degrees Celsius, approximately 113 degrees Fahrenheit, which is markedly higher than those commonly found in conventional data centre cooling environments. Rather than a limitation, Nvidia claims this elevated temperature enhances operational efficiency by diminishing the necessity for energy-heavy cooling apparatus.

Nvidia compared its cooling system to hot tubs, which typically maintain a temperature between 38 and 40 degrees Celsius. The company stated that its latest AI servers can operate with cooling liquid at even higher temperatures. The DSX AI factory reference design is centred on a completely closed-loop liquid cooling system, which continuously reuses coolant without needing fresh water input under normal circumstances. The statements suggest potential for reducing water usage to almost zero, particularly in conditions where dry coolers can manage heat for most of the year.

Ali Heydari, the director of data centre cooling and infrastructure at Nvidia, emphasised that the DSX reference design has achieved zero water consumption for AI factories. He noted that significant power and almost all water usage have been effectively eliminated, with additional cooling systems needed only in exceptional cases dictated by specific climate conditions.

Current Water Usage in Data Centre Operations

While Nvidia outlines its vision for a water-efficient future in AI infrastructure, current practices in cloud operations still depend heavily on water resources. Amazon recently revealed that its global data centres consumed around 2.5 billion gallons, equating to approximately 9.46 billion litres, in a single year. This volume represents about five per cent of the annual water consumption in metro Seattle, where Amazon is headquartered.

The company noted that these figures indicate improvements in efficiency but also vary significantly based on cooling methods, the climate, and the locations of their facilities. Kerry Person, AWS vice president, addressed concerns regarding overall water usage within the industry, suggesting that narratives highlighting excessive consumption are misleading. He argued that detailed analyses of the data reveal a more nuanced picture.

Nvidia’s advancements aim to revolutionise the cooling landscape by enabling reduced energy consumption and lowering long-term operational expenses. The move towards liquid cooling and higher operational temperatures could lead to significant benefits for the data centre sector while lessening the impact on local water resources.

Long or Short, get news the way you like. No ads. No redirections. Download Newspin and Stay Alert, The CSR Journal Mobile app, for fast, crisp, clean updates!

App Store –  https://apps.apple.com/in/app/newspin/id6746449540 

Google Play Store – https://play.google.com/store/apps/details?id=com.inventifweb.newspin&pcampaignid=web_share

Latest News

Popular Videos