Archive notice: This article was originally published on December 14, 2009. Links and embedded videos are preserved as part of the historical record.
The team led by Ting-Wai Chiu, Professor of Physics and Associate Director at the Center for Quantum Science and Engineering (CQSE), studies interactions between subatomic particles to learn more about the origin of the universe. This requires exceptionally high computing power.
NTU uses Taiwan’s first GPU-based supercomputer for this purpose, a 128-GPU cluster at CQSE containing 16 NVIDIA Tesla S1070 systems and 64 Tesla C1060 units. The system is used for quantum-physics calculations.
“Our GPU-based cluster performs significantly better than many conventional supercomputers, both in terms of cost and power consumption,” Chiu emphasises. “The supercomputer with Tesla GPUs provides computing performance of 15 teraflops at a price of USD 200,000. That is only one percent of the cost of a supercomputer such as IBM BlueGene/L.”
Tesla GPUs are based on NVIDIA CUDA technology, a computing architecture that enables developers and researchers to program freely and use the massively parallel computing power of GPUs. CQSE has used it to develop highly efficient, CUDA-accelerated software for complex calculations in quantum physics and astrophysics.
