Skip to main content

HPC Creates Sustainability

sustainability

High-performance computing (HPC) occupies a unique place in the conversation about climate change and sustainability. 

Computational modeling, scientific research, and cleantech innovation are necessary for mitigating and adapting to climate change. At the same time, supercomputing is an energy-intensive business. 

Data centers use tons of electricity for everything from powering the hardware to cooling it down. Artificial intelligence (AI) workloads in particular are notoriously power-hungry. And that’s not even taking into account the carbon footprint that goes into manufacturing hardware. From raw material extraction to assembly to transportation, it all adds up.

The HPC community is tackling this issue head-on. Just walking around SC23, it was obvious that this is a big topic right now. The only word with more prominence than “sustainable” in the exhibit hall was the other topic du jour: AI.

On its own, HPC is not a climate solution. Rather, it comes down to us: the people using and building these supercomputers. Businesses, scientists, and engineers are finding creative ways to make HPC more sustainable and use it as a tool in finding a path to a sustainable future.

Climate Modeling & Scientific Research

Let’s start with the researchers who are taking advantage of today’s most powerful computers to understand the climate situation and search for solutions.

The first angle here is that climate itself and how it may change over time are massive simulation problems with tons of variables and uncertainties. Researchers are using HPC to simulate thousands of scenarios by accounting for possible developments, such as new cleantech hitting the market. 

“The most important contribution of our research is that it allows policymakers to make concrete decisions about climate action based on a full understanding of the existing uncertainties,” Dr. Brian Ó Gallachóir from University College Cork told HPCwire.

tree

Likewise, researchers are also using HPC to engineer climate solutions, including time on the world’s first known exascale supercomputer, the Department of Energy’s (DOE) Frontier machine at Oak Ridge National Laboratory (ORNL). 

The exascale small modular reactor project, or ExaSMR, models the physics involved in nuclear reactors to improve fuel rod design for efficiency and safety. “We used to only be able to simulate the fuel rod at the beginning,” explains Dr. Lori Diachin, Project Director for the DOE’s Exascale Computing Project. “Now we can see the evolution throughout.” 

There are massive sustainability implications for next-generation reactors. ExaSMR was nominated for a 2023 Gordon Bell Prize, which honors outstanding progress in HPC. 

But it’s also far from the only HPC project that’s focusing on sustainability. ExaWind focuses on wind power plant simulations, while other projects focus on carbon capture, creating more fuel-efficient jet engines, and chemical catalysts that reduce excess carbon or other toxic chemicals.

Location, Location, Location

There are hundreds of decisions that go into building a supercomputer and optimizing it for sustainability. But for many, the most basic question with the biggest implications is, “Where do I put this thing?”

For many organizations, it may mean constructing data centers in regions with access to clean energy. Bulk builds infrastructure in Nordic countries with plentiful sustainable electricity, while Crusoe brings the data center to energy sources to use energy that would otherwise be wasted, either from natural gas flaring or excess renewable sources.

Another unique and sustainable approach comes from the Massachusetts Green High Performance Computing Center (MGHPCC), an intercollegiate facility that promotes collaboration between educational institutions in New England. They chose the location so their data center could largely run off hydroelectric power from the nearby Connecticut River.

hydroelectric dam
leed certification

At the same time, location is only part of the equation. Everything from minimizing chiller usage to hot aisle containment is part of the green design that led to the MGHPCC becoming the first university research data center to become LEED platinum certified.

By designing for sustainability from the onset, HPC infrastructure developers can drastically reduce their environmental impact.

Role of Energy-Efficient Hardward

Specific design choices in hardware and how they’re put together can also lead to more sustainable HPC systems. The supercomputers on the GREEN500 list are prime examples of systems engineering with sustainability front and center. The list ranks HPC systems according to their energy efficiency, defined as GigaFlops per Watt. 

The current No. 1, announced at last year’s SC23 conference, is Henri at the Flatiron Institute in New York. Between energy-efficient thermal control, smaller transistor sizes on the NVIDIA chipset, and other optimizations, Henri’s architects have reduced the supercomputer’s energy consumption while maximizing performance. 

The GREEN500 shows that sustainability doesn’t have to come at the expense of performance. Frontier, the system at the top of the TOP500 list, ranked No. 8 on the green list, while the ARM-based Fugaku supercomputer debuted at the  No. 1 on the TOP500 in 2020 while still earning a spot in the top 10 on the GREEN500.

green hardware

Innovative engineering promises to take HPC sustainability one step further. Optical technologies, such as Lightmatter’s photonic integrated circuits and Ayar Labs’ optical I/O interconnects, were standouts at SC23. By swapping out electricity for light, these hardware developers promise to both reduce carbon footprint and improve performance.

Beyond the computing hardware itself, there’s also sustainable innovation happening in the cooling space. Keeping hardware at a safe operating temperature, whether through HVAC or liquid immersion cooling, can use just as much energy – if not more – than the HPC system itself. That’s why innovators like Maxwell Labs are introducing radiative cooling technologies that transform heat into light. 

As long as that light has a path to the sky, it goes straight into outer space without heating the atmosphere. “High performance radiative cooling technology will be the core cooling technology deployed in supercomputing systems in the next decade,” says Jacob Balma, Maxwell Lab’s CEO and Principal Scientist. “We’re enabling novel energy recovery systems and future hardware that is more power-dense than we can imagine today.”

Creating a Sustainable Future

If there’s one thing we know for certain, science and technology are going to be essential in the fight against climate change. And supercomputers are an important tool for enabling scientific and technological progress. 

As we continue to progress further into the AI era, we’re going to see more possibilities open up and additional breakthroughs in areas like cleantech and green energy production.

supercomputer

At the same time, we can’t neglect that running a supercomputer has a real carbon cost. The supercomputing community must come together to advance sustainability solutions – every decision, from more energy-efficient hardware to picking the right location for data centers, has an impact.

We can’t wait to see what sustainability developments will be on display at SC24. Discover them yourself – registration opens July 10.


John Himes is the founder of Dynamic Tech Media, a content marketing agency that translates tech-speak into plain English for HPC, quantum, and other complex technology companies.

Stay Up to Date

Sign up to receive the SC newsletter in your inbox.

Information provided is treated in accordance with ACM & IEEE privacy policies.

Back To Top Button