GPU-ACCELERATED SUPERCOMPUTING SURGES IN RUSSIAOctober 1, 2012
Nearly one-third of the top 50 supercomputers in Russia, including the two fastest systems, get most of their computing capability from NVIDIA Tesla GPUs, according to the latest Top50 Russian supercomputers list from the Russian Supercomputing Consortium.
Russia has been making large investments in high performance computing to fuel the growth of its economy through better science and technology. GPU computing has been growing right along with it.
In fact, just two years ago there were no GPU-accelerated systems among the country’s top 50 supercomputers, as shown in the chart below.
Over the last two years, GPU-accelerated systems have been the only category to consistently grow, while other categories have steady lost ground. Other highlights from the new list include:
- The top two systems, at Moscow State University and Russian Academy of Sciences, are GPU accelerated
- Five of the top 10 supercomputers – and 15 of the top 50 – are GPU accelerated
- Leading OEMs are IBM, HP and T-Platforms
The use of GPUs to accelerate science and engineering is widespread in Russia. Most major universities and institutes in the country teach parallel programing with CUDA GPUs and scientific research programs around GPUs. The scientific applications accelerated by GPUs range from cancer research to space exploration to aircraft engine design.
Computing has become the third pillar in scientific discovery, along with theory and experimentation. By democratizing supercomputing, GPU-accelerated computing has helped countries such as Russia, India, Brazil and China catch up to the U.S. and Europe, because even small GPU systems deliver the performance of leading supercomputers in these regions.