China has claimed the throne among supercomputers as has completed and presented the new supercomputer Tianhe-1A during HPC 2010. The system has reached peak performance of 2507 petaflops, which is largely thanks to the 7000 Fermi Tesla GPUs.
Tianhe-1A was built by National University of Defense Technology (NUDT) and is already up at the National Supercomputer Center in Tianjin. The supercomputer uses thousands of bladeservers housing a total of 14,336 processors and 7,168 Tesla M2050 GPUs, where NVIDIA claims that its graphics circuits is the key to the performance record.
Tianhe-1A epitomizes modern heterogeneous computing by coupling massively parallel GPUs with multi-core CPUs, enabling significant achievements in performance, size and power. The system uses 7,168 NVIDIA® Tesla™ M2050 GPUs and 14,336 CPUs; it would require more than 50,000 CPUs and twice as much floor space to deliver the same performance using CPUs alone.
According to the graphics card maker an equivalent system without Tesla GPUs would consume more than 12 megawatt, but its GPU solutions has lowered this to the actual 4,04 megawatt. A positive note for NVIDIA’s otherwise very power hungry Fermi architecture.
Guangming Liu, chief of National Supercomputer Center in Tianjin, points out that the performance would not be possible without NVIDIA’s Tesla GPUs. Using NVIDIA’s CUDA framework the supercomputer will work with large scale scientific calculations and we eagerly await the next of top500 in a week.