Nvidia has announced
a major win in the world of high-performance computing, with its Fermi-based Tesla GPUs forming the heart of what is now the world's fastest supercomputer.
The Tianhe-1A, created by the National University of Defense Technology in China, uses a combination of 7,168 Nvidia Fermi-based Tesla M2050 graphics processing units and 14,366 Intel processors of unknown type to hit more than 2.5 petaflops, a significant boost over the current 'official' fastest system, the Cray-manufactured Jaguar which measures 1.75 petaflops.
Nvidia claims that in order to reach the same level of performance with CPUs alone, it would require more than 50,000 processors and double the floor space along with a whopping 12 megawatts of power, compared to Tianhe-1A's modest requirement of 4.04 megawatts.
Guangming Liu, head of the National Supercomputer Centre in Tianjin, said of the new system 'the performance and efficiency of Tianhe-1A was simply not possible without GPUs,' stating that 'the scientific research that is now possible with a system of this scale is almost without limits; we could not be more pleased with the results.'
Jen-Hsun Huang, Nvidia's chief executive, claimed that the new system proves that 'GPUs are redefining high performance computing,' and that 'GPU supercomputers are essential tools for scientists looking to turbocharge their rate of discovery.'
The Tianhe-1A will be offered as an open-access system to researchers looking to boost the speed of their scientific modelling and investation across China. Let us know your thoughts on this computing breakthrough in the forums