For something like VGA it can make sense to have a ferrite for signal integrity because the signals inside are not differential and are very sensitive to noise due to being analog. So if you can reduce some common mode crap going over the grounds you can get a actual signal integrity improvement.
I mean... it depends. Usually the cable is triple coax inside braid, so the signal quality can be quite good indeed. The coaxes may go through bead(s) on their way down to the video board -- more the case for classic CRTs (Trinitrons and such) I think, than for single-board digital panels (connector direct into PCB). There's definitely the possibility or opportunity there, to treat it differentially, and reap the improvement in signal quality -- but also one which may be left out by many, and which is rapidly degraded by connectors (DE-15 does not have coax pins on it; they should've never let DB13W3 go
), or discarded entirely (i.e., a novice designer might simply see all the ground pins and think, I'll just tie all these to ground plane and not even think about CMCs).
Cable quality itself varies widely. In the last years I had a CRT, I no longer had a VGA card at all, and had to use a DVI adapter; the stub length and impedance mismatch (or connector quality, or..) introduced a reflection, making a subtle after-image to the right of sharp vertical details (so, most noticable on window outlines for example).
It's all very much more precise and critical than a full-digital or coded scheme is, and for HDMI it's pretty much there or it's blank. So it's no accident we've moved away from [lightly/un-coded] analog signaling in many places.
But anyway, the differential comment -- just to emphasize that better for those that don't know about the techniques -- you can, and absolutely should, treat coax differentially, where possible.
The key point is, you're after the signal between core and shield, so read that signal in such a way that CM currents do not affect it. For example, tying shield to circuit ground through an impedance (there's inevitably some series impedance, shielding is rarely perfect), but terminating the signal to circuit ground, returns that signal current through the series impedance, and thus CM ingress (and emission) occurs. You can terminate the signal to the shield first, so that it flows in a local loop free of CM interference. Then tie shield to circuit ground, so that CM current is sunk separately. The signal voltage still rides on top of that CM noise voltage, it's superimposed upon it. So you use a differential amplifier, or CM filtering, or isolation transformer, or whatever suits the signal, to subtract out the CM error and get clean undisturbed signal referenced to circuit ground. This works anywhere from DC to RF, given that we may want different ground-return impedance in those cases (high at LF, where shielding effectiveness is low, to avoid introducing voltage drop on the shield and therefore the signal reference too; low at RF, for CMRR).
And really, it makes sense, coax is unbalanced to its surroundings, so too should its impedance be unbalanced. The signal line should never terminate towards the surroundings, but only towards its immediate reference plane -- inner shield. Shield outside, common mode, can terminate with whatever impedance, but probably a low impedance to shunt CM current away from the circuit.
Tim