As High-Performance Computing (HPC) has grown, traditional air-cooling methods have become ineffective at handling heat fluxes over 500 W/cm², which has caused performance throttling and reliability problems. This study focuses on the development and computational evaluation of a liquid cooling system model specifically designed for high-performance computing (HPC) environments. A weighted selection matrix showed that single-phase immersion cooling was the best method; it worked better thermally, it could be used on a larger scale, and it used less energy. In Autodesk Inventor, a simplified geometric model of a CPU, GPU, and PCB submerged in a dielectric fluid tank was created. In ANSYS Fluent, the model was then simulated. The thermal performance of four dielectric fluids, 3M Novec 7000, Fluorinert FC-72, Mineral Oil, and Silicone Oil, was evaluated under a constant heat flux of 400,000 W/m³. The results show that 3M Novec 7000 and Fluorinert FC-72 had the lowest component temperatures, which shows that they are very stable at transferring heat. However, Mineral Oil had the best heat absorption capacity (22.52 kJ/kg) and was the cheapest option (ZAR 123.50/L), making it perfect for large-scale HPC deployments. The study confirms the hypothesis that simulated liquid cooling can provide efficient, cost-effective thermal management, with implications for South Africa's HPC sector and global sustainability.

Keywords: High-Performance Computing (HPC), Liquid Cooling, Single-Phase Immersion Cooling, Dielectric Fluids, Thermal Management, ANSYS Fluent, Energy Efficiency, Cost-Effectiveness.