NVIDIA (CEO Jensen Huang), a leader in artificial intelligence (AI) computing technology, has been recognized for its capabilities as an industry leader, with its technology used in more than 70% of the world's top 500 supercomputers.

NVIDIA announced on the 19th that 355 systems, accounting for 70% of the Top 500 list of supercomputers worldwide announced at the recent Supercomputing Conference 2021 (SC21), are being accelerated by NVIDIA technology.

It was also reported that more than 90% of the newly built systems adopt NVIDIA's technology.

23 of the top 25 Green500 systems, which select the most energy-efficient systems, are powered by NVIDIA technology. On average, NVIDIA GPU-based supercomputers provide 3.5 times higher energy efficiency compared to Green500 systems that do not use GPUs.

Microsoft's GPU-accelerated Azure supercomputer ranked 10th, becoming the first cloud-based system to enter the top 10. AI is revolutionizing computing for scientific research. Recently, the number of papers utilizing high-performance computing (HPC) and machine learning has surged, with the number of related papers submitted increasing from about 600 in 2018 to 5,000 in 2020.

HPL-AI is a new benchmark for HPC and AI convergence workloads that utilizes mixed-precision computing (the foundation of deep learning, various scientific research, and commercial applications) while fully delivering the accuracy of double-precision computing (serving as the standard metric for traditional HPC benchmarks).


MLPerf HPC evaluates computing styles that accelerate and improve simulations on supercomputers using AI. It measures performance based on the HPC Center's major workloads: astrophysics (Cosmoflow), weather (Deepcam), and molecular dynamics (Opencatalyst).

NVIDIA covers the full stack with GPU-accelerated processing, smart networking, GPU-optimized applications, and libraries supporting the convergence of AI and HPC. Through this approach, it has been possible to accelerate workloads and achieve scientific innovation.

In various use cases, the combination of GPU parallel processing capabilities and over 2,500 GPU-optimized applications can reduce the time required for HPC tasks from weeks to hours. Because NVIDIA continuously optimizes CUDA-X libraries and GPU-accelerated applications, it is not uncommon to experience unpredictable but powerful performance enhancements on the same GPU architecture.

As a result, the performance of the so-called golden suite, the most widely used scientific applications, has improved by more than 16 times over the past six years, and further development is expected in the future.