NUS ramps up high-performance computing capability to propel research
The rapid advancement of Artificial Intelligence (AI) technologies is transforming research across disciplines—enabling faster data analysis, deeper insights, and new methods of discovery in both foundational AI and AI-driven scientific research. For universities such as NUS, this presents a strategic opportunity to lead in innovation and address complex real-world challenges with greater impact.
To support a growing demand for advanced GPU (Graphics Processing Unit) resources, NUS has expanded its capability by implementing a state-of-the-art supercomputer, Hopper, that can perform a staggering 25 quadrillion calculations per second — or 25 PetaFLOPS — giving NUS researchers access to High-Performance Computing (HPC) power. To put that in perspective, a researcher performing one calculation per second would take millions of years to match what Hopper does in just one second.
The Hopper has been placed in 105th position on the TOP500, a list ranking the world’s most powerful computing systems. Hopper is the highest-ranked supercomputer in Singapore and the only university-operated supercomputer in Southeast Asia to make the cut — marking a significant milestone for NUS and the region in HPC resources for academia.
Driving breakthroughs across disciplines
Since going live in June this year, Hopper has already transformed research workflows across the University, powering ambitious, data-intensive research projects hitherto constrained by time or computing power.
Assistant Professor Mike Shou from the Department of Electrical and Computer Engineering in the College of Design and Engineering (CDE) at NUS described Hopper as a game-changer in his research into developing new deep learning methods to allow machines to understand actions and complex events in videos.
“Its powerful performance and system stability have laid a solid foundation for our work in video generative AI,” he said. “We can now generate both short-form and long-form videos and build multimodal agents that operate across a wide range of previously unseen environments.”
Asst Prof Shou’s work can power many applications such as perception system for self-driving cars — which allow vehicles to sense and interpret their surroundings, healthcare robots assisting patients in wards or the elderly at home, smart CCTV cameras, social media recommendation systems, and intelligent video creation tools for journalists and filmmakers.
Hopper is also driving advances in the field of biomedical engineering. Recently, Assistant Professor Jin Yueming and her team from the Department of Biomedical Engineering at NUS CDE leveraged Hopper to develop SurgVLM, a vision-language model designed to support surgical intelligence.
“Hopper’s NVIDIA H100 GPU nodes and the high-speed InfiniBand network have been instrumental to our key experiments. Once deployed, SurgVLM can provide real-time cognitive assistance during surgery, immersive surgical training, and automation in robotic procedures,” said Asst Prof Jin.
Meanwhile, at the Department of Materials Science and Engineering in NUS CDE, Hopper has helped fast-track next-generation battery design for clean energy infrastructure. Assistant Professor Deng Zeyu and his team have been using the system to run large-scale quantum mechanical calculations and train sophisticated machine learning models to aid the design. With Hopper’s GPU acceleration, they can now complete molecular dynamics simulations and density functional theory calculations up to 12 times faster than with traditional CPU (Central Processing Unit) systems.
“Tasks that used to take three months now take a week,” Asst Prof Deng said. “It’s like switching from a bicycle to a high-speed train — Hopper has completely transformed our research pace and possibilities, allowing us to explore new frontiers in battery technology and support the development of next-generation electric vehicles and renewable energy solutions.”
Levelling the research playing field
Hopper’s shared, centralised infrastructure means that even smaller research teams — those without the resources to maintain their own GPU clusters — can now leverage advanced computing power. From AI and engineering to medicine, climate science, and the social sciences, Hopper is enabling researchers across NUS to scale their work faster and tackle more complex challenges.
By the end of 2025, the system is expected to support up to 120 active research projects. A university-wide call-for-projects was launched to invite NUS researchers from all disciplines to apply for access to Hopper’s GPU capabilities, with onboarding already underway.
Built for the Future
Hopper was designed and deployed by the HPC-AI team at NUS Information Technology (NUS IT) and built with infrastructure from Dell Technologies. Purpose-built to support the university’s growing computational needs, the system emphasises performance, scalability and long-term sustainability.
As research becomes increasingly complex and compute-intensive, Hopper positions NUS to stay ahead — enabling researchers to tackle the world’s pressing challenges with speed and precision.
“This milestone marks the beginning of a new chapter,” said Mr Rikky W. Purbojati, HPC-AI Team Head at NUS IT. “With Hopper, NUS researchers can dream bigger, advance faster, and solve the complex problems of our time.”
With its debut on the TOP500, Hopper affirms NUS’ commitment to research excellence, cross-disciplinary collaboration and building the technological foundations that will power tomorrow’s discoveries.