Linear Iterative Solver + Elliptic PDE
 

When solving numerically for an elliptic PDE (Heat Conduction ) which has huge source terms, should the accuracy of iterative solver be affected by the magnitude of source terms in comparison to the discretized coefficient ap? Even though central differencing schemes insures diagonal dominance, I see the numerical errors increasing with increase in source terms for same levels of convergence. I'm using AMG + GMRES for solving this problem.