New Thermal Material Could Slash Data Center Cooling Demands
Meeting the world’s data storage demands is costly, in terms of money, energy, and environmental impact – but a new material could significantly improve the cooling of our data centers while also making our home and business electronics more energy efficient.
Currently, bulky and energy-intensive cooling solutions are typically deployed to chill out the hardware holding our data, adding up to about 40 percent of overall data center energy use (around 8 terawatt-hours every year).
The team from the University of Texas at Austin and Sichuan University in China estimates around 13 percent of those 8 terawatt-hours could be shaved off by their new organic thermal interface material (TIM).
The TIM substantially boosts the rate at which heat can be taken away from active electronic components and channeled into a heatsink for air or water to carry away.
That in turn means a lower demand on active cooling technologies, including fans and liquid cooling.
Thermal materials aid heat dissipation by moving heat away from the electronic components. (Wu et al., Nature Nanotechnology, 2024)
“The power consumption of cooling infrastructure for energy-intensive data centers and other large electronic systems is skyrocketing,” says materials scientist Guihua Yu, from the University of Texas at Austin.
“That trend isn’t dissipating anytime soon, so it’s critical to develop new ways, like the material we’ve created, for efficient and sustainable cooling of devices operating at kilowatt levels and even higher power.”
The TIM developed here is a colloidal mixture of the liquid metal galinstan and particles of aluminum nitride, combined in a way that creates a gradient interface – one that helps heat pass through without any hard boundaries between the two substances.
Galinstan and aluminum nitride were combined to make the material. (Wu et al., Nature Nanotechnology, 2024)
In an experimental lab test setup, the TIM was able to double the amount of heat that could be safely transferred away from every square centimeter of an electronic component, compared to a leading thermal paste – while also reducing the component’s overall temperature.
The setup used a cooling pump, which is a common protection against overheating, and the TIM cut the energy use of the pump by 65 percent. This was only a small-scale example, but it shows the heat-transferring potential of the material.
“This breakthrough brings us closer to achieving the ideal performance predicted by theory, enabling more sustainable cooling solutions for high-power electronics,” says Kai Wu, from Sichuan University.
The next step is to get the material working on larger systems and in a wider variety of scenarios, something the researchers are already in the process of doing by partnering up with data center providers.
Analysts expect data center electricity usage in 2028 to be double what it was in 2023, driven largely by the increasing demands of artificial intelligence models. That presents a real energy demand problem – one that scientists are working hard to solve.
“Our material can enable sustainable cooling in energy-intensive applications, from data centers to aerospace, paving the way for more efficient and eco-friendly technologies,” says Wu.
The research has been published in Nature Nanotechnology.
See more here sciencealert.com
Header image: Cockrell School Of Engineering
Please Donate Below To Support Our Ongoing Work To Defend The Scientific Method
PRINCIPIA SCIENTIFIC INTERNATIONAL, legally registered in the UK as a company 
incorporated for charitable purposes. Head Office: 27 Old Gloucester Street, London WC1N 3AX.
Trackback from your site.
Howdy
| #
It’s ok saying the TIM moves double the heat, but that will inevitably lead to increased component density, smaller component packages etc, and the end problem is still the same. That is, what does one actually do with the heat to disperse it. That problem won’t go away. It will be exacerbated.
https://www.vertiv.com/en-emea/about/news-and-insights/articles/educational-articles/a-beginners-guide-to-data-center-cooling-systems/
Reply
D. Boss
| #
Ridiculous scam to get more funding for apparently discovered unobtainium. “If only we could afford to scale it up”, is a common lure for grant or investment funding…. The premise is false, as the thermal paste or interface for modern CPU and GPU cores are already not the bottleneck, and damned efficient. The problem is getting the heat out of the intensely dense inner regions of the semiconductor die fast enough. And you cannot use this TIM inside the body of the units if it has any conductivity which I suspect liquid metal has.
I have an RTX 4090 GPU with liquid cooling and it generates 250-450 watts of heat when used at 60-90% of it’s capacity. The present thermal paste on the chips to water block keeps the temps at 45-47 deg C with ease. the pump and fans only draw maybe 20 watts to accomplish this. So this article’s premise is full of schist.
Conversely my same system has an AMD Ryzen 7800x3D CPU also with liquid cooling. The CPU cooler is double the capacity of the GPU, and uses the same thermal paste, but it will run at 60C while only at 20% capacity and rapidly rises to 85C if you load the CPU to 75%. This while it only draws 25-65 watts. The reason is the 7800x3D has very lousy internal heat transfer due to the 3D cache design. Apparently the newer generation 9800’s have solved this internal thermal issue.
But new thermal paste is not going to solve the issue of really dense chip design getting the heat out of the silicon structure.
Reply