South Korea to Launch 6th National Supercomputer with 614 PFLOPS in December
South Korea's Ministry of Science and ICT announced that the country's 6th national supercomputer, built in partnership with Hewlett Packard Enterprise (HPE), will officially launch in December 2024. Featuring 8,496 GPUs—including NVIDIA GH200 Grace Hopper superchips—and 9,936 CPUs, the new system delivers a theoretical peak compute performance of 614 PFLOPS, which is 24 times powerful than its predecessor. The massive upgrade aims to propel South Korea into the top 10 of the global TOP500 supercomputer rankings, strengthening its national infrastructure for AI and advanced scientific research. By allocating 55% of compute capacity to general research and 15% to enterprise initiatives, the system will significantly lower HPC resource barriers for both academia and industry. Beyond its 614 PFLOPS raw compute performance, the supercomputer features 205 petabytes of storage and network transfer speeds of at least 400 Gbps, representing a 10-fold storage increase and a 4-fold networking boost over the 5th generation. The system is currently undergoing performance and stability testing following the completion of physical installation and basic checks.
## BACKGROUND
Supercomputer performance is commonly measured in FLOPS (floating-point operations per second), where 1 PFLOPS (petaFLOPS) represents one quadrillion operations per second. Modern AI and scientific HPC clusters rely heavily on specialized hardware like the NVIDIA GH200 Grace Hopper superchip, which integrates an Arm-based Grace CPU and a Hopper GPU on a single module using high-speed NVLink-C2C interconnects to accelerate data-intensive workloads.