A computational campaign was carried out at the Department of Astronautical, Electrical and Energy Engineering of Sapienza University of Rome aiming at the assessment of RELAP5-3D© capabilities for subchannel analysis. More specifically, the investigation involved a lead-bismuth-eutectic–cooled wire-spaced fuel pin bundle and compared simulation outcomes with experimental data coming from the NAtural CIrculation Experiment-Upgraded (NACIE-UP) facility, hosted at ENEA Brasimone Research Center. Thermal-hydraulic nodalization of the facility was developed with detailed subchannel modeling of the fuel pin simulator (FPS). Three different methodologies for the subchannel simulation were investigated, increasing step by step the complexity of the thermal-hydraulic model. In the simplest approach, the subchannels were modeled one by one. In addition, mass transfer between them was considered thanks to multiple cross junction components, realizing the hydraulic connection between adjacent subchannels. In this case, mass transfer depends on the pressure gradient and hydraulic resistance only, ignoring the turbulent mixing promoted by the wire-wrapped subassembly. Simulation results were not satisfactory, and an improvement was introduced in the second approach. In this case, several control variables calculate at each time step the energy transfer between adjacent control volumes associated with the turbulent mixing induced by the wires. This energy is transferred using ad hoc heat structures (HSs), where the boundary conditions are calculated by the control variables. The present model highlighted good capabilities in the prediction of the radial temperature distribution within the FPS, considerably reducing disagreement with experimental data. Finally, the influence of radial conduction within the fluid domain was assessed, introducing further HSs. Although this most complex model provided the best estimation of the experimental acquisition, the improvements given by radial conduction were not so relevant to justify the correspondent increase of the computational cost.