ASC25 finalists announced as universities compete in high-stakes global supercomputing challenge
The ASC Student Supercomputer Challenge (ASC25) has selected 25 university teams to advance to the finals, following a competitive preliminary round involving more than 300 institutions worldwide.
The competition, recognized as the world’s largest student supercomputing event, brings together university teams to tackle complex high-performance computing (HPC) and AI challenges. With the goal of fostering innovation in scientific computing, system optimization, and real-world applications, ASC has grown into a key platform for developing young talent in the field.
The 2025 finals, hosted by Qinghai University from May 10–14, will see leading institutions such as Zhejiang University, Peking University, Tsinghua University, University of the Witwatersrand, and Friedrich-Alexander-University Erlangen-Nuremberg competing for top honors, including the Champion, Silver Prize, Group Competition Award, e Prize, and Highest LINPACK.
Strong performances in the preliminary round
Among the top performers, Zhejiang University took the lead in the preliminaries, demonstrating advanced optimization strategies and precise performance analysis. Other returning finalists include Peking University, last year’s ASC champion, and Sun Yat-sen University, which finished as runner-up in ASC24. Established competitors from Tsinghua University, Shanghai Jiao Tong University, and Friedrich-Alexander-University Erlangen-Nuremberg will also contend for major awards.
This year’s competition also features several new finalists, including Nanjing University, Beijing Normal University, and the University of the Witwatersrand, adding an element of unpredictability to the upcoming finals.
Technical challenges in ASC25
The preliminary round required teams to solve two complex computing challenges:
RNA Methylation Modification Site Detection: Participants optimized the detection of 5-methylcytosine (m5C) sites on mRNA, working with three datasets containing 340 million sequencing reads. The challenge tested their ability to streamline workflow steps, configure software environments, and maximize computational performance.
AlphaFold3 Inference Optimization: Teams adapted AlphaFold3’s protein structure prediction model for a CPU-based platform, optimizing inference efficiency while maintaining prediction accuracy.
Breakthroughs in supercomputing optimization
Several teams delivered standout performances in computational efficiency and workflow optimization:
Zhejiang University achieved an 11× speed increase by implementing data parallelism for the hisat-3n-table application. They also rewrote Java-based programs in C++, enhancing memory efficiency and input/output (I/O) performance, leading to a 6.6× speedup.
University of the Witwatersrand focused on parallel execution, distributing datasets across multiple computing nodes with a cluster scheduler, achieving a 3× speedup. They further improved I/O performance by using RAM disks for temporary workflow files.
Peking University delivered the best performance in AlphaFold3 optimization, converting the original JAX-based code to PyTorch, applying operator fusion, data type conversion, and AMX utilization, and increasing inference speed by 14.5×.
Kasetsart University addressed NaN errors in AlphaFold3’s denoising process and optimized CPU core selection to enhance computational efficiency while maintaining prediction accuracy.
Finals set to showcase next-generation supercomputing talent
With a mix of experienced contenders and first-time finalists, the ASC25 finals are expected to be one of the most competitive in the event’s history. Teams will refine their HPC and AI strategies as they compete for top honors and demonstrate innovations in scientific computing and optimization.
Since its launch in 2012, the ASC Student Supercomputer Challenge has attracted university teams from six continents, making it the largest university-level supercomputing competition in the world. The competition is supported by leading experts and institutions across Asia, Europe, and America, with a focus on advancing research, fostering global collaboration, and pushing the boundaries of high-performance computing and AI integration.
