Four of the top twenty-five supercomputers in India featured on the Top Supercomputers – India list released by the Indian Institute of Science (IISC) are powered by NVIDIA Tesla GPUs to accelerate and improve the output of high performance computational design and analysis.
ISRO’s SAGA, which has recently also been featured at the 86th position on the global Top500 Supercomputers List, along with clusters from Punjab Agricultural University (PAU), Ludhiana, Indian Institute of Science (IISc), Bangalore, and Indian Institute of Technology (IIT), Chennai, are amongst the Top 25 supercomputers in India. The four supercomputers utilize the strength of NVIDIA Tesla 20 series GPUs (Graphics Processing Units). NVIDIA CUDA computing platform and programming model allows scientists and researchers to harness the power of GPUs for improved performance in parallel computing, helping to build a High-Performance Computing Environment (HPCE). With OpenACC directive based programming, end users get to work on a standard designed to further simplify parallel programming of heterogeneous CPU/GPU systems.
NVIDIA GPUs implement hundreds of simple, power-efficient cores that are optimized for high throughput on parallel workloads. Hybrid architectures work to improve application performance per watt by shifting most of the work to the throughput-optimized cores, while using the CPU cores for the residual serial work. Since a core cannot be optimized for both energy-efficiency and fast single-thread performance, the hybrid architecture allows concentrating on making the GPU cores more and more energy efficient, while relying on the CPU cores for serial performance.
“GPU-based supercomputing systems have already been adopted in leading economies such as the US, China, Europe. India is also recognizing the potential that can be unlocked in various industries by harnessing the power of a GPU now,” said Vishal Dhupar, Managing Director – Asia South, NVIDIA.
