A Performance Comparison of x86 and Arm on OpenFOAM for Computational Fluid Dynamics

As the importance of High-Performance Computing is continuously grows, the Arm architecture emerges as a powerful alternative to the conventional x86 architecture, driven by its advantages in power efficiency and core scalability. In this st udy, the performance of two different CPU architectures, x86 and Arm, was empirically compared and validated to determine their competitiveness in practical CFD applications. The x86 processor has 16 cores, whereas the Arm processor has 128 cores. For the analysis, single workstation environments for both x86 and Arm architectures were configured, and the open-source Computational Fluid Dynamics software, OpenFOAM, was utilized to assess the performance of these two CPU architectures. Two fluid analysis cases, incompressible and compressible flow, were selected. To ensure a fair comparison, the numerical analyses were performed on both architectures using the same number of CPU cores. The incompressible flow was evaluated using 4 to 32 cores, and the compressible flow was tested from 16 to 32 cores. The performance of each architecture was compared and analyzed based on the actual computation time. In terms of computational time, reduced times were demonstrated by the Arm architecture compared to the x86 architecture. For the incompressible flow analysis, computation time of the Arm process was, on average, 14% lower on the coarse mesh and 24% lower on the fine mesh. A similar trend was observed in the compressible flow analysis, with average computational time reduction of 12% on the coarse mesh and 21% on the fine mesh. In conclusion, this study demonstrates that the Arm processor offers better core scalability, cost-effectiveness, and power efficiency over the x86 processor. Furthermore, a comparison of the computational fluid dynamics analysis time confirms that Arm processor performance is competitive enough to be an alternative to the x86 processor.