Super Micro Computer and Preferred Networks (PFN) earned the top spot for the Green500 semi-annual industry assessment, with their MN-3 (Deep Learning Supercomputer MN-Core) achieving a record 21.11Gigaflops of performance-per-watt on a benchmark run that delivered a total performance of 1.62 Petaflops.
This efficiency achievement is 15% higher than the previous Green500 record of 18.404 Gflops/W, which was recorded in June 2018.
“Supermicro was excited to work with PFN on this exceptional system supporting machine learning and deep learning applications and energy efficiency,” said Charles Liang, CEO and president of Supermicro.
“The Green500 international recognition confirms that Supermicro delivers exceptional products – resource-saving, superior design, and high-reliability to the market,” said Liang.
PFN partnered with Supermicro to develop this customised server, which addresses a wide range of applications that require ultra-fast communications. Based in Japan, PFN strives to solve complex problems that use deep learning, robotics, and other advanced technologies.
“We are very pleased to have partnered with Supermicro, who worked with us very closely to build MN-3, which was recognised as the world’s most energy-efficiency supercomputer,” stated Yusuke Doi, VP of computing infrastructure at Preferred Networks. “We can deliver outstanding performance while using a fraction of the power that was previously required for such a large supercomputer.”
The PFN solution is based on the Supermicro GPU server that uses Intel Xeon CPUs, and MN-Core boards developed by Preferred Networks. This advanced system offers up to 6TB of DDR4 memory, and multiple GPUs or accelerators, as well as the interconnects that enable ultra-fast communications between GPUs.
The PFN supercomputer can serve many users simultaneously. The initial cluster consisted of 48 servers — four interconnect nodes — and five 100GbE switches. There was a total of 2,080 CPU cores, housed in a 7U high rack-mounted unit.