China may be secretly possessing two of the most powerful supercomputers in the world

According to a new report from Next Platform, China has built machines that can break the “exascale barrier” in closed testing processes. The source shared in the report said that a device at the National Supercomputing Center (also known as Sunway Oceanlite) located in Wuxi City, Jiangsu Province, scored the highest score of 1, 3 exaFLOPS (by Linpack standards) in early March this year.

Another system, Tianhe-3, is said to have achieved almost the same score, but it is unclear exactly when the test took place. Although little is known about the architecture of the machine in Wuxi, Tianhe-3 is known to have a design based on chips developed by a Chinese company called Phytium, with an architecture developed by the Chinese company. driven by a matrix accelerator.

“Exascale” is a term to refer to the computer system’s ability to process at least 1 exaFLOPS, which is 1 billion billion operations per second. Thus, if the above information is correct, both Chinese supercomputers have broken this barrier.

However, both are not on the list of Top 500 Global Supercomputers. The reason is said to be because their developers don’t want their partner’s subcontractors to get in trouble with the US government.

Tianhe-2A, China’s most recently announced supercomputer.

The record for the world’s fastest supercomputer is currently held by a Japanese machine, Fugaku. It earned this top spot in June 2020 with a score of 416 petaFLOP (or 0.416 exaFLOP), nearly three times the peak performance of the previous leader, IBM Summit.

Since then, Fugaku’s lead has been solidified with the addition of 330,000 cores, bringing performance to 442 petaFLOPS. However, if the reports are correct, both Tianhe-3 and the Sunway Oceanlite engine are almost three times ahead of the current leader.

The arrival of exascale supercomputers is expected to open up a wide range of opportunities in many different fields. For example, this level of performance will accelerate discovery and research times in areas such as clinical medicine and genomics, which require vast amounts of computing power to perform molecular modeling and genome sequencing. gene.

Artificial intelligence (AI) is another multidisciplinary field that will also receive transformational steps with the arrival of exascale computing. The ability to analyze increasingly large datasets will improve the capabilities of AI models, enabling accurate forecasts that can be applied in virtually any context, from cybersecurity to commerce. electronics, manufacturing, logistics, banking and more.

The Frontier supercomputer isn’t out yet, but it’s backed by many big American companies.

As both the US and China are vying for supremacy in the field of AI, the emergence of two exascale-capable systems in China before the US can launch its own upcoming exascale machine (named Frontier), would be a blow to the Biden administration. Even more special when they are built on China’s own technology platform.

NextPlatform said it got its information from a source from the US who knew what was going on in China.

However, there are some factors to consider. The Frontier supercomputer’s target performance at launch is around 1.5 TFLOPS, almost 50% higher when compared to the sustained performance of China’s exascale supercomputers. Furthermore, Frontier is expected to consume about 30 MW of electricity, while one of its rivals in China has an electricity consumption of about 35 MW. Last but not least, Chinese developers use existing architectures developed for PetaFLOPS-scale systems and workloads, and this is a standard that may not be optimal. Future.

synthetic