Evaluating multi-GPU computing capabilities of Numba and CuPy

Askar T. Lukac M. Shukirgaliyev B. Abdikamalov E.
October 2025 Springer

Cluster Computing
2025 #28 Issue 11

In this paper, we evaluate the performance of Numba and CuPy in multi-GPU configurations, focusing on both strong and weak scalings. We employ two benchmark problems: pseudo-random number generation and one-dimensional Monte Carlo radiation transport in a purely absorbing medium. In the experiments, we compare Numba, CuPy, and CUDA C implementations under both default and optimized conditions implemented on the NVIDIA DGX-2 server platform. In the default setup, CUDA C delivers better performance and the highest energy efficiency. However, we demonstrate that CuPy can achieve substantial speedups in optimized mode, though this requires extensive code modifications. Numba shows competitive performance in cases with minimal data transfer to global memory, but its scalability and energy efficiency are limited by CPU-side random number generator state initialization bottlenecks, and it benefits less from optimizations compared to CuPy.

CUDA , CuPy , Multi-GPU , Numba , Performance

Text of the article Перейти на текст статьи

School of Engineering and Digital Sciences, Nazarbayev University, 53 Qabanbay Batyr Ave, Astana, 010000, Kazakhstan
Department of Computer and Network Engineering, Hiroshima City University, Ozukahigashi 3-4-1, Hiroshima, 731-3166, Japan
Department of Physics, Nazarbayev University, 53 Qabanbay Batyr Ave, Astana, 010000, Kazakhstan
Heriot-Watt International Faculty, K. Zhubanov Aktobe Regional University, 34 A. Moldagulova Ave., Aktobe, 030000, Kazakhstan
Energetic Cosmos Laboratory, Nazarbayev University, 53 Qabanbay Batyr Ave, Astana, 010000, Kazakhstan

School of Engineering and Digital Sciences
Department of Computer and Network Engineering
Department of Physics
Heriot-Watt International Faculty
Energetic Cosmos Laboratory

10 лет помогаем публиковать статьи Международный издатель

Книга Публикация научной статьи Волощук 2026 Book Publication of a scientific article 2026