May 11, 2025
Trending News

Hot water, cool servers: Nvidia’s Blackwell chips as pool heaters

  • November 7, 2024
  • 0

Lenovo says hot water is the solution to keeping servers cool. With some sensible innovations, the manufacturer wants to make water cooling accessible across the board. HPC and

Lenovo says hot water is the solution to keeping servers cool. With some sensible innovations, the manufacturer wants to make water cooling accessible across the board. HPC and AI are the initial goals, but in the long term everyone can benefit from them. The key: a solution that fits into existing data centers.

Air cooling of servers is reaching its limits. That’s what Scott Tease, vice president of WW High-Performance Computing and AI Intelligence at Lenovo, said in an interview with ITdaily during Tech World in Bellevue. “All the energy you pump into a server leaves the device in the form of heat,” he explains.

“If your server needs ten kilowatts of power, you have to dissipate ten kilowatts of heat,” he continues. “This usually happens with air conditioning, but if you want to save 10 kW, you need about 4 kW of cooling.”

If you want to remove 10 kW of heat, you need about 4 kW of cooling.

Scott Tease, vice president of WW High-Performance Computing and AI Intelligence, Lenovo

It’s not efficient, but it works. “Even HPC servers with a consumption of 25 kW can be cooled with air,” admits Tease. Lenovo itself offers such solutions and is exhibiting them at Tech World. The air-cooled HPC servers are huge and weigh a lot. The huge radiators play a big role here.

You can still cool AI servers with air today, but the huge coolers take up a lot of space. In addition, the servers are extremely heavy, so a rack must quickly be placed on a reinforced base.

Air cooling limit

As servers become more powerful, the air cooling option is disappearing. “You simply can’t cool 100 kW with air,” says Tease. The solution; Water cooling.

You simply can’t cool 100 kW with air

Scott Tease, vice president of WW High-Performance Computing and AI Intelligence, Lenovo

Lenovo isn’t the only one who has recognized this. “Dell and HP also develop powerful solutions,” says Tease himself. “But their shelves are huge. They are 8 feet deep, won’t fit through a door or fit in an elevator, and customers will need to reinforce their data center floor to install them.”

Accessibility issue

The trick, therefore, is to cool powerful servers in a more accessible way. “Our water-cooled servers are also slightly wider and taller than a standard rack server,” said Tease. In other words, they don’t fit in a standard rack either.

“We presented this challenge to the engineering team,” he continues. “How can we fit these oversized servers into a standard 19-inch rack?” At Tech World, Lenovo is presenting the solution to this problem.

Lenovo’s N1380 chassis fits into a 19-inch rack and accommodates vertically installed, water-cooled servers.

The manufacturer introduces the N1380 chassis. This is a large box in which customers can store HPC servers vertically. The N1380 chassis houses the servers vertically and fits in a classic 19-inch rack. The housing also has PDUs (Power distribution units) on board. One rack can hold three N1380 speakers. With this, Lenovo opens the door to racks with more than 100 kW of power without the need for drastic adjustments within the data center.

Hot water

The Neptune liquid cooling itself works with uncooled water at room temperature. “The water entering the server can be up to 45°C,” says Tease. Water is created with a temperature of approx. 55 °C. Tease: “This means we no longer have to invest energy in cooling the water itself.”

With these temperature tolerances, the sixth generation of Neptune liquid cooling is in principle suitable for keeping most data centers in Europe and the USA fresh without any special pre-treatment of the water.

Thanks to years of in-house research, Lenovo is able to offer this fairly unique technology. This research actually began in IBMtime and is based on findings from the mainframe area. “If you look at the first solutions developed around 2010, you can see that Neptune is still based on them today,” says Tease. “Large pipes, minimal connections, low flow and low pressure: the solution still follows the same principles.”

Lenovo designed Neptune from the ground up, ensuring water draws heat directly away from the chip for maximum efficiency. This server has two Nvidia Blackwell Superchips on which the GPU and CPU complexes are cooled directly.

Lenovo designs its water-cooled servers from the ground up, starting with the motherboard. According to Tease, this is the key to efficiency, which in turn enables hot water cooling.

Swimming pool heating

It goes without saying that users have to connect liquid cooling, but according to Tease, this doesn’t pose a big challenge. Lenovo has extensive experience in integrating water cooling into customer data centers that were originally built for air cooling.

Tease hopes that customers will use the residual heat in the future, for example to heat buildings. This isn’t a theoretical idea: Lenovo already heats meeting rooms at its factory in Hungary with hot water from its Neptune test servers.

“You can simply connect Neptune’s water lines to pipes and pump the water somewhere else,” he explains. “Imagine, for example, a university where the warm water from the data center can heat the water in the swimming pool via a heat exchanger. The hot water is no longer a waste product, but can be recycled to save energy.”

Low PUE

The integration of the water cooling itself also brings a lot of savings. “It all starts with the fans,” says Tease. “Fans account for ten to fifteen percent of the power consumption of air-cooled servers. It takes a lot of energy to move air.”

With Neptune-cooled servers, they no longer need fans, and because water cooling removes all residual heat from the servers, the data center no longer needs air conditioning.

“This way we save a huge amount of electricity,” says Tease. “An air-cooled data center typically has one Effectiveness of power consumption (PUE) of 1.4 or 1.6, which means that forty to sixty percent of the energy goes not to the servers, but to the cooling. With Neptune the PUE drops below 1.1. A PUE of 1.07 is easily achievable.”

In other words, water cooling helps save up to 40 percent power compared to the same workload supported by air cooling.

AI has really accelerated interest in water cooling.

Scott Tease, vice president of WW High-Performance Computing and AI Intelligence, Lenovo

“We initially developed the technology for HPC customers,” explains Tease. “But AI will now be the biggest beneficiary. AI has really accelerated the interest in water cooling.”

Also for companies?

The numbers Tease uses also seem interesting outside of AI and HPC workloads. More classic servers can also benefit from water cooling. A lower PUE means lower environmental impact and lower operating costs.

“Interest is actually greater than ever,” agrees Tease, “but not yet as strong from outside of HPC and AI.” With sustainability goals and ESG targets, there are more motivating factors than ever to choose water cooling, but we do not yet see a major upswing in the corporate segment. Not yet, because that will happen in the long term.”

Sexy servers

At Tech World, Lenovo is introducing the brand new SC777: an AI server cooled with Neptune water cooling and equipped with two Nvidia GB200 Blackwell superchips. CEO Yuanqing Yan presents the server on stage together with Nvidia boss Jensen Huang. “Isn’t that beautiful,” Huang said at the introduction. “For an engineer, this device is sexy.”

For an engineer, this device is sexy.

Jensen Huang, CEO Nvidia

Nvidia CEO Jensen Huang calls the water-cooled Lenovo SC777 sexy.

“This SC777 uses fifteen times more power than the first nodes we cooled with Neptune about fifteen years ago,” adds Tease. “That seems like a lot, but each SC777 is about a thousand times more efficient than the servers of its time. The devices use more, but the performance per watt has improved exponentially.

Relatively accessible

Neptune and the N1380 chassis make the SC777 relatively affordable. Thanks to Neptune, customers can deploy classic racks with the AI ​​servers in a classic data center. Of course, this doesn’t mean that everyone can suddenly install Blackwell AI servers, not to mention the price of things.

Tease: “For a long time, the limiting factor for the use of sufficient computing power was the budget and the number of computing cores. Today that is actually the available power.” Lenovo can no longer feed electricity into the grid. Accessible Neptune solutions must ensure that the power available to a business is used as much as possible for IT hardware and not cooling.

Secret sauce

Tease believes Neptune will give Lenovo an advantage. “Other manufacturers can probably copy our solution if they really want to,” he suspects. “But that doesn’t mean that they have our experience with converting traditional data centers to water cooling. This experience is our secret recipe.”

Source: IT Daily

Leave a Reply

Your email address will not be published. Required fields are marked *

Exit mobile version