A "magnetic dipole" is, as described above, a vague and ambiguous term. Indeed, magnetic dipoles do not even exist in nature (div(B) = 0, always). In a general sense, though, all magnetic field lines are not always normal to the current-carrying surface.
This follows directly from the Biot-Savart law, in the general case of a surface current:
B(r) = [itex]\frac{\mu}{4\pi}[/itex] [itex]\int[/itex] [itex]\frac{K(\acute{r}) χ \hat{r}}{r^2}[/itex]d[itex]\hat{\tau}[/itex]
where K([itex]\acute{r}[/itex]) is the surface current density,
and [itex]\hat{r}[/itex] is the vector extending from the source to the point r
We note that the direction of the magnetic field will be given by the cross product between a vector pointing in the direction of the current and a vector pointing towards the point. Ergo, the magnetic field lines must always be perpendicular to the direction of current, but may not be perpendicular to the surface itself.
In your particular case, the field lines will always be perpendicular to the outer edges ("dipoles") of the bar.
Hope this helped. :3