InfiniBand Technology Continues to Power Majority of High Performance Computing Systems on TOP500 List
The InfiniBand Trade Association (IBTA), a global organization dedicated to maintaining and furthering the InfiniBand specification, shared the latest TOP500 List results that report InfiniBand accelerates the top three supercomputers in the world and six of the top 10 systems. Additionally, HDR 200Gb/s InfiniBand, the fastest end to end interconnect technology currently available, makes its debut on the List in four systems including the fifth highest ranking supercomputer, the Frontera at Texas Advanced Computing Center.
“This milestone is in line with the InfiniBand Roadmap, an industry resource that provides clear timelines around expected InfiniBand performance improvements that will ensure investments made in HPC and AI systems will be protected.”
Other TOP500 List Highlights:
- InfiniBand continues to be the interconnect of choice for High Performance Computing (HPC) systems, connecting 126 supercomputers, or nearly 55 percent of the overall systems in the HPC category.
- InfiniBand accelerates the top two fastest industrial supercomputers, the PANGEA III system at Total Exploration Production, ranked number 11 on the List, and the HPC4 system at Eni S.p.A., ranked number 17 on the List.
- 170 RoCE-capable Ethernet systems listed, representing a majority of all Ethernet systems.
“We are excited by HDR 200Gb/s InfiniBand’s debut on the latest Top500 List, which once again highlights InfiniBand technology as the high-performance interconnect of choice for HPC and an increasing amount of AI platforms,” said Bill Lee, IBTA Marketing Working Group Co-Chair. “This milestone is in line with the InfiniBand Roadmap, an industry resource that provides clear timelines around expected InfiniBand performance improvements that will ensure investments made in HPC and AI systems will be protected.”
The TOP500 List has evolved to include both traditional HPC and enterprise platforms, with more than half of the platforms on the TOP500 List being categorized as cloud/hyperscale application platforms. HPC capabilities have also increasingly been used for Artificial Intelligence (AI) in both research and commercial facilities. Applications such as AI and Machine Learning (ML) require an extraordinary amount of compute power and low latency that Remote Direct Memory Access (RDMA) fabrics provide in order to address scalability and efficiency needs for ever-growing, complex data sets. RDMA fabrics such as RDMA over Converged Ethernet (RoCE) and InfiniBand are used to power these demanding AI applications being deployed today.
The TOP500 List is published twice per year and ranks the top supercomputers worldwide based on the LINPAC benchmark rating system, providing valuable statistics for tracking trends in system performance and architecture.