Atrial fibrillation (AF), the most common arrhythmia, can be diagnosed from altered morphology of P-waves in patient's ECG. However, in AF P-waves are absent and small, erratic fibrillatory waves (f-waves) can be observed. In-silico modelling of cardiac electrophysiology can provide mechanistic insights into AF-related ECG changes, and it typically relies on solving bidomain equations. Yet, this approach is computationally expensive and challenging to scale. This study evaluates alternative methodologies for forward calculations of ECG: comparing f-waves generated using the monodomain model and the infinite volume conductor method against the pseudo-bidomain is used to evaluate the relative accuracy of the approaches. For the pseudo-bidomain approach, we used a 3D torso model generated from high-resolution CT patient scans, incorporating the entire heart, blood pools, bones, lungs, liver, and kidneys. The monodomain model utilized both volumetric and bilayer bi-atrial representations. In each case, the atria were subdivided into atrial regions based on electrophysiological characteristics, and AF-related ionic remodelling of the atria and patient fibrotic maps from LGE-MRI scans were also incorporated. Finite elements simulations were performed using openCARP software. AF in all models was initiated using an identical rapid pacing protocol from the same location on the left superior pulmonary vein. ECGs were recorded for the duration of rapid pacing and after AF reentrant drivers stabilization near the right inferior pulmonary vein. Additionally, we simulated and compared sinus rhythm (SR) P-waves across all models as a baseline. Analysis of the signal Pearson correlation and dominant frequencies of SR P-wave and AF f-wave ECGs across the different methods demonstrate that the bilayer bi-atrial and volumetric monodomain models, combined with the infinite volume conductor approach, can accurately simulate AF-related ECGs. This methodology offers a computationally efficient alternative (x9 speedup) and can significantly enhance the scalability of in-silico clinical trials.