National Cheng Kung U Develops Switchless Cluster Supercomputer

The National Cheng Kung University (NCKU) Supercomputing Research Center (RSC) in Taiwan has developed a switchless cluster supercomputer, called CK-Star.

CK-Star connects eight Acer Altos R380 F2 servers with Intel Xeon E5-2630 v2 processors and Intel Xeon Phi 7120P coprocessors to achieve 15 teraFLOPS (15 trillion floating-point operations). According to NCKU, the CK-Star "has produced record-breaking computer performance at 80.2 percent for four nodes and 77.3 percent for eight nodes, beating Intel’s previous world record of 79.6 percent for four nodes and 76.1 percent for eight nodes."

Unlike traditional cluster supercomputers, which use switches to control interaction between the computer nodes, the CK-Star uses a switchless design. According to NCKU, switches introduce a performance bottleneck when there are a large number of nodes in the cluster, and they also consume considerable power, as much as 50 percent of the total power consumption of a traditional cluster supercomputer.

"CK-Star has an innovative structure and an enhanced performance. It has overcome the limitations of switches and solved the problem of high power dissipation through switches," said Yuefan Deng, one of the developers of CK-Star, in a prepared statement.

CK-Star is not the first record-breaking supercomputer built by NCKU RSC. The center also collaborated with Gigabyte to develop the GS-R22PHL supercomputer. The GS-R22PHL has been recorded to reach 3.7 teraFLOPS (3.7 trillion floating-point operations) and is the highest performing single supercomputer to date, according to NCKU.

CK-Star was built by Chi-Chuan Hwang, director of NCKU RSC, and Yuefan Deng of mainland China's National Supercomputing Center in Jinan (NSCCJN), in collaboration with an international team of researchers in Asia and the United States. Yuefan Deng is also a distinguished professor at the State University of New York (SUNY).

About the Author

Leila Meyer is a technology writer based in British Columbia. She can be reached at [email protected].

Featured

  • futuristic representation of interconnected individuals within a digital network

    OpenAI Launches Safety Fellowship to Fund External AI Research

    OpenAI is expanding safety efforts beyond its walls with a new Safety Fellowship that will fund external researchers to study AI risks.

  • Abstract energetic glowing lines with particles

    Meta Steps Up Enterprise AI Ambitions with Muse Spark Launch

    Meta has announced the launch of Muse Spark 1.1, a multimodal reasoning model designed for agentic AI, alongside a new Meta Model API that gives developers access to the model for the first time.

  • Abstract neural network 3D illustration

    IntelĀ® AI EmpowerED: The AI-Ready Campus, Delivered

    Artificial intelligence is transforming higher education, prompting institutions to rethink how they manage infrastructure, security, governance, and workforce readiness. Successful adoption requires a strategic, institution-wide approach that aligns AI initiatives with educational goals, faculty enablement, and scalable operational frameworks.

  • clock and neon light trails

    Don't Wait for the Clock to Run Out on Digital Accessibility

    Public universities with over 50,000 students face the looming April 24, 2026, deadline to comply with new Americans with Disabilities Act (ADA) Title II standards. The urgency many feel is warranted: Implementation timelines are tight and the scope of compliance is extensive.