The preliminary results for the 2025 ASC Student Supercomputer Challenge (ASC25) are in! Following a rigorous evaluation of the preliminary results from student teams representing 300 universities worldwide, the ASC Committee has selected 25 outstanding teams to advance to the highly anticipated finals. This year’s finalists include leading institutions such as Zhejiang University, Peking University, Tsinghua University, University of the Witwatersrand, and Friedrich-Alexander-University Erlangen- Nuremberg. These teams will showcase their supercomputing prowess as they compete for top honors, including the Champion, Silver Prize, Group Competition Award, e Prize, and Highest LINPACK. Hosted by Qinghai University from May 10 to 14, 2025, the ASC25 finals promise to be a thrilling showcase of innovation, teamwork, and cutting-edge computing in pushing the frontiers of AI and HPC integration.
Among the finalists, Zhejiang University stood out with an exceptional display of supercomputing expertise. Their team demonstrated a deep understanding of high-performance computing through precise performance bottleneck analysis and innovative optimization strategies, achieving remarkable results. Their outstanding performance across all tasks secured them the top rank in the preliminary round, setting the stage for an exciting showdown at the finals.
Among the top contenders are traditional supercomputing powerhouses, including Peking University—champion of the ASC24 finals, Sun Yat-sen University—last year’s runner-up, as well as Tsinghua University, Shanghai Jiao Tong University, and Friedrich-Alexander-University Erlangen- Nuremberg. Adding to the excitement, this year’s competition welcomes first-time finalists, including teams from Nanjing University, Beijing Normal University, and University of the Witwatersrand. Their debut on the ASC stage brings fresh energy to the competition, promising a dynamic clash between established leaders and rising challengers. With a mix of seasoned champions and ambitious newcomers, the ASC25 finals are set to be one of the most competitive and unpredictable yet.
The ASC25 preliminaries tested the skills of student teams with two advanced challenges: RNA methylation modification site detection and AlphaFold3 inference optimization. These tasks highlighted the integration of high-performance computing (HPC) and artificial intelligence (AI) in groundbreaking scientific research. To tackle these challenges, teams conducted in-depth research on industry-leading optimization algorithms, devising innovative solutions tailored to their specific software and hardware environments. By maximizing computational performance, they showcased exceptional learning abilities, practical problem-solving skills, and expertise in supercomputing application analysis, performance optimization, and parallel strategy design.
For the RNA task, teams were required to optimize the detection process of 5-methylcytosine (m5C) modification sites on mRNA. This involved streamlining workflow steps, configuring the software environment, and efficiently processing three datasets containing approximately 340 million sequencing reads. The objective was to maximize workflow performance while maintaining high accuracy, pushing teams to develop efficient computational strategies in bioinformatics.
Zhejiang University showcased remarkable technical expertise and innovative thinking, implementing a data parallelism strategy to reconstruct the core application hisat-3n-table, achieving an 11-fold performance improvement. They also rewrote a Java program in C++, enhancing memory efficiency and I/O performance, leading to a 6.6-fold speedup. These two key optimizations resulted in an outstanding overall performance in workflow execution. The University of the Witwatersrand focused on parallelizing workflow execution. By analyzing steps that could run concurrently, they leveraged a cluster scheduler to distribute three datasets across multiple computing nodes, achieving a 3x speedup compared to sequential execution. Additionally, they optimized I/O performance by upgrading their storage system and utilizing RAM disks for temporary workflow files, significantly improving execution efficiency. With these breakthroughs, both teams delivered commendable performance in the RNA methylation detection challenge.
The AlphaFold3 inference optimization task required participants to optimize the structure prediction code of AlphaFold3 by migrating it from a GPU-based platform to a CPU-based system. Their objective was to implement efficient inference for 12 protein sequences designated by the organizing committee, while maintaining high prediction accuracy and demonstrating their ability to optimize deep learning-based protein structure prediction on CPUs.
Among the standout performers, Peking University achieved the best performance on the task of CPU optimization. They successfully ran the original JAX AlphaFold3 code on CPU and re-implemented the model in PyTorch based on xFold. With Pytorch, they optimized the inference performance via operator fusion, optimization of data type conversion, and AMX utilization, achieving an impressive 14.5× inference speed increase. Kasetsart University also tackled key AlphaFold3 code efficiency challenges. They first addressed a NaN issue during the denoising process when running AlphaFold3 on CPU, and made appropriate choice of the number of CPU cores based on sequence length. They then optimized XLA compiler flags and used FP32 precision, reducing inference time while maintaining prediction accuracy. With these impressive optimizations, both teams showcased cutting-edge problem-solving and computational efficiency.
About ASC:
The ASC Student Supercomputer Challenge serves as a platform to promote the exchange and development of young talent in supercomputing worldwide, with support from experts and institutions across Asia, Europe, and America. The competition aims to elevate the application and R&D capabilities in supercomputing, harness its technological driving force, and foster innovation across science, technology, and industry. Since its inception in 2012, ASC has attracted tens of thousands of university students from six continents, solidifying its status as the largest university-level supercomputing competition globally. To discover more about this impactful endeavor, visit the website http://www.asc-events.net/StudentChallenge/index.html
View source version on businesswire.com: https://www.businesswire.com/news/home/20250318678781/en/