Supercomputer on the third generation of Loongson processors

Last week, the world's media spread the news that China announced the construction of a Dawning 6000 supercomputer with a capacity of more than 1000 teraflops on its own Loongson (aka Godson) processors.

The launch of the supercomputer will take place this summer and it will surely enter the Top 20 in the world.

Until now, Loongson processors were used only in weak netbooks and set-top boxes. Since they can emulate x86-code, the question arises: does the third generation of these chips already have enough power to compete with Intel processors?

The Loongson processor family is being developed at the Chinese Academy of Sciences . The first models were introduced in 2002 and since then they have been gradually improved, but much remains to be done. The chief architect of the Loongson project and a member of the Chinese parliament, Hu Weiwu, said in an interview with People's Daily Online that it takes about ten years before Loongson processors begin to meet the requirements of local consumers, and another decade to become competitive in the international market. Only twenty years later, China will be able to export its own processors to the US and other countries "just as it exports clothing and sneakers," said Hu Waive.
')
According to Wave, the Chinese IT industry cannot depend on foreign technology. The development of supercomputers on their own CPUs is one of the priority national projects in the field of science and technology. Three organizations independently create their own supercomputers - the Institute of Computer Technologies at the Chinese Academy of Sciences, the Jiangnan Institute of Computer Technologies and the National University of Defense Technologies (NUDT).

The Dawning 6000 supercomputer, which is being created jointly by the Chinese Academy of Sciences and Dawning Information Industry, will be the first project to use Loongson processors. It is assumed that it will have less than 10,000 microchips and the system will be much more energy efficient than its competitors.

For comparison, in the world's most powerful supercomputer Tianhe-1A, which in October 2010 launched the National University of Defense Technologies, employs 14,336 Intel chips and 7,186 Nvidia chips, two American companies, which is not good.

The Chinese have not yet developed enough software applications to use this computing power, so the existing supercomputers are now loaded by no more than 10%. According to Hu Waew, the reason for this is the lack of programmers who specialize in this area and can develop the necessary algorithms for data processing. This is very sad, because the annual bill for the power supply of each data center is about $ 1.5 million.