AI-powered models have the ability to monitor greenhouse gases and predict extreme weather events, as demonstrated by a joint effort by IBM and NASA. However, there is still a concern: What is the environmental impact of maintaining these energy-intensive models?
The environmental impact of data centers depends on factors such as power and water consumption, as well as the lifespan of the equipment. Recently, a new tool developed by MIT students helps reduce energy consumption by 80% to train AI models.
The United States is already home to some of the fastest supercomputers worldwide, a fact corroborated by TOP500 rankings.
These computational marvels serve a variety of purposes, including applications in the defense sector, as reported recently when the Los Alamos National Laboratory (LANL) revealed plans to employ a supercomputer for monitoring nuclear stockpiles for the U.S. military.
In stark contrast, Aurora has a unique mission: to optimize the nuclear fission processes within reactors that currently provide a significant portion of the United States’ electrical power, amounting to a fifth of the nation’s supply.
Additionally, nuclear energy accounts for nearly half of the country’s carbon-free electricity, contributing significantly to efforts to reduce carbon emissions.
But how powerful is the Aurora supercomputer exactly? The ANL currently employs the Polaris supercomputer for simulations, boasting a formidable 44 petaflops of computing power, capable of performing 44 quadrillion calculations per second.
In contrast, Aurora is designed with a mind-boggling two exaflops of computational capability, allowing it to complete an astounding two quintillion calculations per second, a monumental leap in performance compared to the existing system.
Although Aurora was initially expected to be operational by this point, manufacturing issues have delayed its completion. However, when it becomes operational, it will surpass Oak Ridge National Laboratory’s Frontier, securing its position as the world’s fastest supercomputer.
As Dillon Shaver, a nuclear engineer at ANL, stated in the press release, “What’s really new with Aurora are both the scale of the simulations we’re going to be able to do and the number of simulations.”
So, what exactly will Aurora be used for? Shaver and his team aim to harness Aurora’s computational prowess to tackle the countless unknowns within nuclear reactor simulations.
Their goal is to capture intricate details of processes taking place inside the reactor core, enabling the development of new reactor designs without the need for costly experiments. This, in turn, will aid reactor manufacturers in validating and licensing their designs.
In their simulations, researchers will scrutinize the complex patterns of heat, such as turbulence, surrounding fuel pins to model the heat transfer properties of the reactor. Increasing turbulence can enhance heat transfer, but it also demands more energy.
Particularly in sodium-cooled fast reactors, turbulence can give rise to vortices, which are small heat whirlpools that can intensify and lead to fuel pin vibration. The research team will utilize fluid dynamics in conjunction with structural mechanics and fuel performance in their simulations, all at the exascale (The fastest supercomputers in the world today solve problems at the petascale—that is a quadrillion (1015) calculations each second) level.
To facilitate these simulations, the team will employ the Multiphysics Object Oriented Simulation Environment (MOOSE), making the modeling and simulation accessible to numerous researchers.
While MOOSE accelerates the simulation process, Aurora’s computational might allows it to work in tandem with NekRS, a computational fluid dynamics solver, enabling the simulation of even the minutest details.
As Shaver emphasized, “These small-scale dynamics are really important to tease out because they compound together to give you the large-scale behavior of the heat transport in the reactor.
We have to get as fundamental as we can to make sure we get the best answers that we can.” Aurora’s impending capabilities mark a significant leap forward in the field of nuclear reactor simulation, with the potential to revolutionize reactor design and safety.