Bishop Caroline
June 10, 2025 01:37
NVIDIA’s accelerated computing technology promotes modern supercomputers to a new height, enabling breakthroughs of scientific research and efficiency shown in ISC 2025.
Modern high -performance computing (HPC) transcends its traditional role in rapid calculations to supply fuel to artificial intelligence (AI) systems that unlock significant scientific breakthroughs. According to a recent article by NVIDIA, the company is at the forefront of evolution and accounts for 77%of the system in the latest TOP500 list of the world’s most powerful supercomputers in ISC 2025.
NVIDIA dominates the TOP500 and Green500 lists
The influence of NVIDIA in the supercomputing stadium is emphasized by the 381 system in the TOP500 list, including Jülich SuperComputing Center, which debuted in #4. In particular, 83 of the top 100 systems are currently using accelerated computing and only 17 depend on the CPU. In addition, NVIDIA’s GH200 Grace Hopper SUPERCHIPS emphasizes the energy efficiency by emphasizing the top three systems in the Green500 list.
Tensor Core: AI innovation and scientific discovery
The center of the influence of NVIDIA is tensor core to quickly carry the matrix work essential for AI and deep learning. These components allow faster calculations, especially as organizations switch to lower precisions such as FP8 for model education. NVIDIA is actively developing libraries to improve high -precision tensors and matrix calculations, which focus on accuracy, performance and energy efficiency.
Yuki Uchino and Katsuhisa Ozaki’s studies showed that there is a possibility of using the integer matrix multiplication for any precision, including FP64. Inspired by this, NVIDIA is creating a library that has already shown significant advantages such as 1.8 times the speed of silicone simulation and preserving time and energy.
AI SuperComputing: Scientific Progress Leading
In supercomputing, mixed precision and the ability of AI are leading to scientific development. Notable achievements include the David Keyes team for their work on the Nobel Prize and ERA5 climate data sets awarded to researchers who use AI to predict protein structure.
The Isambard-Ai system of Bristol University, driven by NVIDIA Grace Hopper, shows the use of mixed precision in educational models such as Nightingale, a multimodal basic model for medical and biological research. This model shows the potential of AI in developing a scientific workflow by providing medical insights by integrating various data types.
HPC’s future
The integration of accelerated computing, tensor technology, and mixed precision methods reorganizes computational science to open a way for AI -centered innovation. The future of supercomputing, such as Jupiter -like system and Ozaki emulation, is ready for deformed development.
As the field develops, the center will satisfy the needs of science and HPC communities worldwide with a smart and flexible approach that does not damage integrity and prioritizes scientific discoveries.
For more information, visit the NVIDIA blog.
Image Source: Shutter Stock