There is lot of information on the net about Allen deviation charts,
But on the short, on such chart can see how a oscilator performs for frequency stability on the short and on the long run.
GPS is on the short time very bad, due to radio path variations and reflections and number of satelites in view.
That can be as bad as 10E-6. But on the long term GPS is very very good, because it follows the ground stations corrections.
In an allen dev chart you can see this performance over time.
A temp controlled oscilator is very good on the short term, but less on the long run.
So if you combine these two, GPS and a TCO you get the best of these.
I have two lucent rubidium oscilators, and a HP5370B counter which i can use to compare with the Bodnar GPSDO.
My GPSDO locks in about 2 to 3 seconds, and is then better than 10E-9
After 10 minutes i get about < 3 E-10 measured with the HP5370, interval ~1 second. ( see graph of the next 10 minutes = 400 samples)
The rubidium clocks are about <5 E-11
And about the magnetic feet , thats is only for easy use on metal, like cars