NHN Cloud said on the 26th that the core GPU cluster of its artificial intelligence (AI) infrastructure brand "NHN FactoryX" ranked highest in Korea in the recently released "Global Supercomputer TOP500."
Global Supercomputer TOP500 is a representative high-performance computing (HPC) ranking that has announced the performance of supercomputers worldwide since 1993 through the International Supercomputing Conference (ISC) in Germany and the Supercomputing Conference (SC) in the United States. The 67th ranking was released this time.
NHN Cloud was selected last year as a key operator for the Ministry of Science and ICT's "GPU procurement, construction, and operation support project," securing 7,656 Nvidia B200 GPUs. Based on this, it built the "NIPA-CL1" GPU cluster with 4,080 units (510 nodes) and the "NIPA-CL2" GPU cluster with 2,040 units (255 nodes).
In benchmark results, NIPA-CL1 recorded a measured peak performance (Rmax) of 137.4 petaflops (PF), placing No. 20 globally and No. 1 in Korea. NIPA-CL2 posted 68.42 PF, ranking No. 40 globally and No. 4 in Korea. With these results, NHN Cloud now operates two of the top five supercomputers in Korea.
The company said it improved GPU cluster efficiency based on GPU infrastructure operation experience accumulated since 2023 by building and running the Gwangju national AI data center. In particular, NIPA-CL1 is the largest GPU cluster in Korea based on the Nvidia B200 and features an InfiniBand network and a liquid-cooling system.
At "FactoryX Seoul," an AI-dedicated data center where the GPU clusters are installed, a 100% liquid-cooling GPU system that supports a high-density environment of 75 kilowatts (kW) per rack was applied. The company said this improved infrastructure stability compared with conventional air cooling.
NHN Cloud also published the "NHN FactoryX technical white paper," which contains its experience in building and operating GPU clusters. The white paper covers AI-dedicated data center architecture and GPU cluster configuration, liquid-cooled data center design, GPU as a service (GPUaaS) operating structure, and cases of building private AI infrastructure for corporations.
Kim Dong-hoon, CEO of NHN Cloud, said, "This achievement not only shows that NHN Cloud is working at the forefront of the national AI infrastructure acquisition project, but also that, through outstanding GPU cluster build-out capabilities, it stands as the undisputed No. 1 AI infrastructure operating company in Korea," adding, "We will not keep this result as an internal accomplishment only, but will share it widely and grow together with the domestic AI ecosystem as a corporations."