When dealing with spd matrices of relatively low condition number, how likely is it (and is it easily provable) that the conjugate gradient method will always be able to find the solution without requiring any kind of preconditioning?

How stable are many of the common preconditioners such as JACOBI, MIC, AINV and is there any way of knowing which are likely to fail to converge if any?