The SAGA system, India’s most powerful supercomputer and holder of the 86th position on the Top500 list released yesterday, employs NVIDIA GPUs to improve the design and analysis of the delivery vehicles critical to the nation’s space program. Developed by the Indian Space Research Organization (ISRO), one of the six largest space research agencies in the world, the SAGA supercomputer is used to tackle complex aeronautical problems.
Harnessing the power of 640 NVIDIA Tesla GPUs and providing up to 394 teraflops of peak performance, SAGA enables ISRO to accelerate and improve the design and analysis of new and existing satellite launch vehicles by enabling more complex and accurate design simulations.
“GPUs enabled us to reduce the time to create, simulate and verify a new launch vehicle design from weeks to days, while improving the overall quality and durability of the design dramatically,” said sources at ISRO. “This level of supercomputing power has made a huge difference in our space program, by enabling us to continually refine and improve our prediction capabilities.”
ISRO leverages the SAGA system and the PARAS-3D application to run compute-intensive, detailed computational fluid dynamics (CFD) simulations that optimize the launch vehicle design for improved performance under a range of flow conditions. With GPUs, ISRO is able to run more complex simulations 7-8 times faster than with a CPU-based system, reducing the time to create, test and finalize new designs by the same factor. In addition, the higher compute density in a GPU based system enables better application scalability while requiring a much lower investment in communication infrastructure.
“The Indian government has jumped into global supercomputing race in a big way, with its recent $1 billion pledge to build a network of next-generation, petascale-class supercomputers throughout the country,” said Sumit Gupta, Senior Director, NVIDIA’s Tesla business. “India joins the US, China, Germany, Japan, Russia, and other leading nations who are applying supercomputing resources to solve some of the worlds’ most pressing computational, scientific and engineering problems.”
