Chipmaker Nvidia announced that a new supercomputer built in China, powered by over 7,000 of its graphics processor units (GPUs), is now the world's fastest. This supercomputer, constructed by the National University of Defense Technology and located at the National Supercomputing Center in Tianjin, has a processing power equivalent to 175,000 laptop computers, according to Nvidia officials.
With sustained performance reaching 2.5 petaflops, this Chinese supercomputer, named Tianhe-1A, is 30% faster than the world's second most powerful supercomputer at Oak Ridge National Laboratory in Tennessee, as per Nvidia's claims. The Tianhe-1A will serve scientists in various fields and will also be available for international use.
Nvidia, known for its high-end video cards favored by gamers, is now promoting its technology for diverse applications beyond graphics, including supercomputers that run complex simulations in astrophysics and other computation-heavy tasks. Supercomputers are essential for medical imaging, oil exploration, and weather forecasting.
Parallel Processing Capabilities
Traditional central processors (CPUs) in computers are designed to perform large calculations sequentially at high speed. In contrast, GPUs excel at handling multiple smaller calculations simultaneously, making them ideal for specific tasks.
"If we wanted to grow beyond our traditional gaming market, we needed to expand our product portfolio. That's why we decided a few years ago to invest in making the GPU more programmable," said Ujesh Desai, Nvidia's vice president of product marketing.
The high-end graphics chips used in the Tianhe-1A sell for about $2,500 each. Additionally, the supercomputer includes over 14,000 central processors made by Intel, according to Nvidia.
Currently, seven of the world's most powerful computers are in the US, with two in China and one in Germany, based on Nvidia's data. Some experts argue that because programmers and major semiconductor companies like Intel have traditionally focused on CPUs, GPUs have been underutilized. They believe leveraging GPUs could significantly improve the performance of everyday computers, from desktops to tablets.
Intel and rival AMD both plan to release microchips next year that combine graphics and traditional processing power, aiming to enhance overall computing efficiency.
