One note for DG : square waveforms aren't particulary correct at frequency upper than 1MHz ( transitions are curve !!!). Sinus at 60MHz very correct.
Ι've noticed that as well on my DG1022z; one would think that "upgrading" it to dg1062z would also scale upwards how the square responds for frequencies > 1MHz but it hasn't changed much.
The relationship should be linear since the square is a sum anyways so essentially doubling the BW would also double the limit of a nice square output.
If it was 1meg before, now it should be at least double that.
I've not tested a "before" and "after" to be sure but I too think that square is the same.
With the 25MHz limit an acceptable square output of around 1MHz maybe be expected since a square inherently consists of infinite harmonics to start with and technically *any* BW is not enough to represent it but it's weird that changing the upper BW limit didn't improve the square output much (enough to be noticeable at least)..
EDIT:
I've had a problem (user error probably) after upgrading so for anyone experiencing the same it might be of help.
The peak to peak value was not consistent all the way up to 60MHz.
I had to revert to factory settings in order to have a proper voltage output all the way up to the limit.
But that did not improve the square I think.
Τhen again, the scope also has a BW limit so..
I'd need to change the model back to 1022 and take a diff measurement of some sort to at least have a reference point.