Gordon Bell Finalists Harness AI Supercomputers for Science

Five research teams have been named finalists for the prestigious Gordon Bell Prize, recognized for their groundbreaking work using NVIDIA-powered supercomputers to tackle some of science's most complex s. Announced at the SC25 supercomputing conference, these projects span climate modeling, materials science, fluid dynamics, and geophysics, demonstrating how advanced computing is accelerating across multiple disciplines.

The finalists are leveraging NVIDIA's supercomputing infrastructure, particularly the Alps system at the Swiss National Supercomputing Centre (CSCS), to push computational boundaries that were previously unattainable. Thomas Schulthess, director of CSCS, emphasized that "without the Alps supercomputer, these scientific discoveries simply would not exist," highlighting the critical role of advanced hardware in enabling these breakthroughs.

One standout project comes from researchers at the Max Planck Institute for Meteorology and collaborating institutions, who have developed a novel configuration for the ICON Earth system model. Running at kilometer-scale resolution, ICON can simulate approximately 146 days of Earth's systems in just 24 hours, providing unprecedented detail in climate modeling and weather forecasting. This temporal compression enables more efficient projections of climate patterns decades into the future.

Another finalist, ORBIT-2, represents a collaboration between Oak Ridge National Laboratory and NVIDIA that creates an AI foundation model for weather and climate downscaling. The model overcomes limitations of traditional climate simulations by generating high-resolution data from lower-resolution sources, allowing researchers to capture localized phenomena like urban heat islands and extreme precipitation events with greater precision.

In semiconductor design, ETH Zurich researchers have advanced nanoscale electronic device modeling with QuaTrEx, a package of algorithms running on NVIDIA GH200 Superchips. The system can simulate devices with more than 45,000 atoms, enabling faster and more accurate design of next-generation transistors crucial for the semiconductor industry. Professor Mathieu Luisier noted that access to Alps allowed simulations "we could not imagine handling just a few months ago."

Space exploration also benefits from these computational advances through MFC, an open-source fluid flow solver developed by Georgia Institute of Technology researchers. The system enables rocket engine simulations that run four times faster with over five times greater energy efficiency while maintaining accuracy, potentially revolutionizing spacecraft design by modeling complex engine interactions that previously d engineers.

Perhaps most dramatically, a collaborative effort between The University of Texas at Austin, Lawrence Livermore National Laboratory, and UC San Diego has produced the world's first digital twin capable of issuing real-time probabilistic tsunami forecasts. The system achieved a staggering 10 billion-fold speedup, completing computations that would normally take 50 years in just 0.2 seconds, providing crucial early warning capabilities for coastal communities.

All finalists have made their openly accessible on ArXiv, supporting the broader scientific community's ability to build upon these advancements. The projects collectively demonstrate how the intersection of AI, high-performance computing, and specialized hardware is transforming what's possible in scientific research across multiple critical domains.