Survey of GPU-Accelerated HPC Applications Across Scientific Domains

GPU‑accelerated high‑performance computing (HPC) growth has been almost entirely driven by NVIDIA, which has built a robust software ecosystem—including parallel programming APIs, libraries, and development tools such as CUDA, OpenCL, and OpenACC—to support GPU platforms.

NVIDIA has also established more than 20 GPU centers worldwide and a global network of academic GPU research and education centers, accelerating the development of general‑purpose GPU programming tools and collaborating with ISVs to improve commercial HPC applications and libraries.

With the rise of AI, many organizations are investing in deep‑learning technologies that heavily rely on GPUs, making AI the main growth engine for NVIDIA.

In the "HPC Applications Supporting GPU Computing" survey conducted by consulting firm Intersect360, users listed their top HPC applications; among 1,792 reported programs, 534 were distinct, and 1792 statistical programs were analyzed.

The report focuses on the top 50 HPC applications reported in user site surveys, selecting two applications from each major scenario to keep the list to 50 (including a tie for 49th). GPU‑accelerated applications span many fields, such as chemistry, fluid dynamics, structural analysis, environmental modeling, geophysics, visualization/image processing, and physics.

In the chemistry domain, 20 of the top 50 HPC applications are used, driven by interest in biomolecular research and new compound design; 16 of these already support GPU acceleration.

In fluid dynamics, GPU‑accelerated CFD tools like ANSYS Fluent and OpenFOAM are highlighted.

Structural analysis applications, including explicit and implicit finite‑element analysis (FEA), are also largely GPU‑enabled, with 7 out of 8 listed tools supporting GPUs.

Biological sciences, such as genomics and proteomics, are beginning to adopt GPU versions of tools like GPU‑BLAST and GPU‑accelerated Bowtie.

Other domains covered include weather and environmental modeling, business intelligence (e.g., GPU‑accelerated SAP and Oracle), physics, and pattern recognition, where deep‑learning frameworks like TensorFlow benefit from GPU acceleration.