Nmedia - Fotolia

China's supercomputing systems soar from 21 to 219 in a decade

The U.S. dominates in total supercomputing performance, but lags behind China in the overall number of supercomputers. U.S. x86 vendors sell many of the chips used in these systems, but ARM may soon be giving them a run for their money.

China may soon be home to half of the world's most powerful supercomputing systems.

Ten years ago, China had 21 systems on the Top500 list of the world's largest supercomputing systems. It now has 219, according to the biannual listing, which was updated just this week. At its current pace of development, China may have half of the supercomputing systems on the Top500 list by 2021.

China is using its own chip technology and U.S.-made processors to build the supercomputers. Chips made by Intel and AMD dominate, but ARM processors may soon get more interest, thanks to Nvidia.

This week, Nvidia announced it will now support ARM CPUs. Nvidia GPUs are widely used in supercomputing. Five of the world's top 10 supercomputers are using its technology. GPUs process computationally intensive mathematical problems faster and at a lower cost than CPUs.

"If you want to get a lot of a parallel threads going, ARM is a very good way to do that," said Nathan Brookwood, principal analyst at Insight 64. "Each thread will use less power than a comparable x86 thread."

ARM is a supercomputing contender

ARM is a contender in supercomputing -- and geopolitics may help it, Brookwood said. Those who want an ARM architecture license can build their own custom processor. They can also take advantage of the ARM ecosystem, he said.

There is concern in China that the U.S. may limit x86 chip imports for supercomputing systems, according to Brookwood. The Chinese are working to expand their alternatives, including using the RISC-V processor architecture. But they could also turn to the ARM architecture, "if that looks more promising," he said.

Supercomputing systems are seen as critical to product development, scientific research and national defense. Supercomputers enable scientists to test and develop new technologies and drugs in virtual environments, as well as study complex systems, such as climate and artificial intelligence.

Nvidia believes ARM's use in supercomputing systems is set to expand.

"We expect x86 CPUs to remain dominant in the short term," said Paresh Kharya, director of product marketing at Nvidia, based in Santa Clara, Calif. "But there's growing interest in ARM for supercomputing, as evidenced by projects in the U.S., Europe and Japan."

Kharya continued, saying, "Supercomputing centers want choice in CPU architecture."

ARM processors have a long way to go if they are going to make a dent in the Top500 list. Lenovo has 173 supercomputers represented on this list, making it the top vendor of supercomputers. But all its systems are built on Intel Xeon processors, and many include Nvidia GPUs.

In the recent past, Lenovo released an ARM-based server that was based on the ThunderX ARM processor.

"We saw little customer demand beyond testing and have not continued to roadmap," said Scott Tease, executive director of hyperscale and high-performance computing for the Lenovo enterprise business group.

Japan's Post-K ARM system

Interest in ARM supercomputers is still early. But one notable ARM-based system on the way is coming from Japan.

Fujitsu is building an ARM-based supercomputer. It is the successor to the K system, which was, for a time, the world's fastest supercomputer. The Post-K system is due next year. Fujitsu said it will work with ARM open source communities.

The Top500 also ranks nations by supercomputing performance.

U.S. supercomputers make up 116 of the latest Top500 list. Despite being well behind China in total system count, the U.S. leads in overall performance, as measured by the High Performance Linpack (HPL) benchmark. The HPL benchmark is used to solve linear equations.

The U.S. has about 38% of the aggregate Top500 list performance. China is in second, at nearly 30% of the performance total.  

But this performance metric has flip-flopped between China and the U.S., because it's heavily weighted by the largest systems.  

The U.S. owns the top two spots on the latest Top500 list, thanks to two IBM supercomputers at U.S. national laboratories. These systems, Summit and Sierra, alone, represent 15.6% of the HPL performance measure.

How long will the U.S. maintain the performance lead?

"It could be wiped out in the next list," said Jack Dongarra, one of the academics behind the Top500 list and director of the Innovative Computing Laboratory at the University of Tennessee. Dongarra developed the HPL benchmark.

Next Steps

Japan named HPC leader as world races to exascale

Dig Deeper on Data center design and facilities

Cloud Computing
and ESG