ACADEMIA

NEC Corporation (NEC) announced the purchase of a large-scale supercomputer system consisting of 20 SX-8R vector supercomputers by the Cybermedia Center (CMC) at Osaka University. The new system achieves a peak performance of 5.3 TFLOPS and will be the largest SX series system acquired in Japan. CMC plans to add a next-generation SX system in two years, whose peak performance is expected to exceed 20 TFLOPS (one trillion floating-point operations per second), a performance enhancement of 16 times that of the current system of SX-5/12M8 (peak performance: 1,280GFLOPS). The Cybermedia Center at Osaka University was established in April 2000 and is a joint product of the former Computing Center (CC) and related departments of Osaka University. The center has continually deployed computing systems of the highest levels and provided computational resources for large and high-speed calculations to researchers in a wide range of areas across Japan, resulting in significant achievements in large-scale simulations in multiple and varied fields. CMC has highly evaluated NEC's system for its user-friendly programming realized by high sustained performance and large-scale shared memory, which has been proved on benchmark tests with the user's actual applications. Recently, cluster systems, which consist of many CPUs connected in parallel, have shown a great improvement in performance, achieving well in areas such as molecular science, semiconductor design, structural analysis and bio informatics. However, in the areas of fluid and particle analysis, vector supercomputers continue to demonstrate absolute dominance in terms of CPU performance and calculation efficiency when compared with cluster systems. In addition, efficiency improvement in power consumption and installation make vector supercomputers superior in operation cost to that of cluster systems. NEC launched sale of the world's fastest commercial vector supercomputer SX-8R in October 2006. The SX-8R boasts a peak vector performance of 144 TFLOPS and demonstrates high performance in the areas of automotive performance, weather forecasts and environment simulations, where large-scale, large-capacity data calculations are required.