Logo Die Sachsen News
News / Technology

Supercomputer Discovery: HZDR team beats NASA competition

This is how cancer could be treated in the future. The simulation shows how a laser accelerates protons to high energies. The calculations will be performed on the Discovery supercomputer.
PIConGPU simulation: A laser beam (red) strikes a hydrogen beam and accelerates protons that could be used to treat tumors in the future. © Dr. Richard Pausch/HZDR
From: Wissensland
HZDR researchers from Dresden and Görlitz have been selected as the only non-American team to receive early access to the U.S. supercomputer Discovery. Their AI-powered simulation program, PIConGPU, could help improve cancer treatments and generate clean energy.

Nine projects from around the world will soon be able to conduct research on the future U.S. supercomputer Discovery. Among them is a team from the Helmholtz-Zentrum Dresden-Rossendorf (HZDR). Together with the University of Delaware (USA), the team prevailed over competitors such as NASA and General Electric. It is the only international project to do so.

Behind this success is a program called PIConGPU. It uses computer simulations to model how laser beams interact with plasma. Plasma is an extremely hot, electrically charged gas. Such simulations help develop new technologies. These include compact particle accelerators for cancer treatment and systems for laser-driven nuclear fusion, which could one day provide climate-friendly energy.

More from this category

Where AI Saves on the Most Expensive Bills

What makes PIConGPU special is its AI-supported approach. High-performance computers are expensive and have limited capacity. That’s why artificial intelligence is used to identify in advance which simulations are actually worthwhile. Only the most promising ones are then actually run. This saves computing time and makes research more efficient.

“We are incredibly proud that we’ve also made it into the CAAR program for ‘Discovery,’” says Dr. Michael Bussmann, research team leader at the Center for Advanced Systems Understanding (CASUS) at HZDR in Görlitz. He hopes that the team will succeed for the first time in combining large AI models with high-resolution simulations to determine the correct operating parameters for new technologies.

“Since we’ve expanded PIConGPU to include several AI-based optimizations over the past two years, we had secretly hoped to be selected for the Discovery CAAR program. Nevertheless, we were overjoyed when the news of the award arrived,” adds Prof. Sunita Chandrasekaran of the University of Delaware.

From Görlitz and Dresden to the World’s Top

The Discovery supercomputer is scheduled to be delivered by 2028 at the latest and to go into operation in 2029. It will be significantly more powerful than its predecessor, “Frontier,” which is currently one of the fastest supercomputers in the world. PIConGPU was already running on “Frontier.” With “Discovery,” even larger and more precise simulations will be possible in the future. At the HZDR, the Center for Advanced Systems Understanding (CASUS) in Görlitz and the Institute of Radiation Physics in Dresden are jointly leading the project.

Bussmann also emphasizes the European perspective. The HZDR is part of a European research network that provides access to Europe’s most powerful supercomputers. JUPITER, located at the Jülich Research Center, is one of Europe’s most powerful computers. "It is important that the HZDR is part of the EuroHPC Center of Excellence PlasmaPEPSC. Access to European exascale computers like JUPITER ensures that we remain at the forefront of research."

PIConGPU has been under continuous development for more than ten years. The simulation software is available free of charge to researchers worldwide.

The translations are automated with the help of AI. We look forward to your feedback and your help in improving our multilingual service. Write to us at: language@diesachsen.com.
Wissensland
Article from

Wissensland

Wissensland is responsible for the content itself. The platform's code of conduct applies. The platform checks and treats content in accordance with the legal requirements, in particular the NetzDG.

METIS