Here's an actual example of a character using a greatweapon with and without GWM, and with +2 Strength respectively.
Before the choice point: Strength +3, To hit +6, Base damage 3+2d6=10
After the choice point:
A. Strength +4, To hit +7, Base damage 11
B. Strength +3, To hit +6, Base damage 10 or 20
Versus AC 18:
A. Hit probability 50% (needs to roll 11) = expected damage 11x0,5=5,5
B. Hit probability 45% (needs to roll 12) so does not use feat
Actual hit probability 45% x 10 = expected damage 10x0,45=4,5
Winner A by one point (much as expected)
Versus AC 12:
A. Hit probability 80% (needs to roll 5) = expected damage 11x0,8=8,8
B. Hit probability 75% (needs to roll 6) so uses the feat
Actual hit probability 50% x 20 = expected damage 10
Winner B by not much more than one point.
So far the feat seems balanced or underpowered even.
But this does not take into consideration the numerous ways of boosting your GWM usage.
So let's compute vs AC 12 with advantage, shall we?
Versus AC 12 with advantage:
A. Hit probability 96% (needs to roll 5 once) = expected damage 11x0,96=10,56
B. Hit probability 9375% (needs to roll 6 once) so uses the feat
Actual hit probability 75% x 20 = expected damage 15
Winner B by four and a half point, or over +40% damage.
Before we move over to conclusions, I'm posting this so you can check so I haven't made a mistake, or suggest why another set of comparison points would improve the analysis.
OK - so first, it seems now we agree on what I said earlier, which is that at best (vs AC 12, with advantage, and against a low AC mob) you only get about a 4 average damage increase w/GWM.
As for the 40% increase, I think to be fair, that is misleading. First because any damage buff at a low level when expressed as a +x% is going to sound OP (OMG! my mage just got a +1 dagger and his dps went up 100% - yay!!! sic). How about when you are 20 STR w/a +2 weapon and your avg dmg.is 14? Now its 28%.
And baddies' AC and hp go up in somewhat proportionate patterns. So if the AC is as low as you mention, there hp are going to be lower and your "bigger chunk" damage is more likely to produce overkill. Now I know overkill can be mitigated by cleave to some extent, but it is merely mitigated a bit, not eliminated as a soak to avg damage.
Examples:
vs Orcs w/AC 13 and 15 hp, you win - no likely overkill and low AC - also GWM allows you to 1 shot him - perfect opponent
vs ANKYLOSAURUS w/AC15 and 68 hp - overkill is a dicey affair, and AC weakens GWM's effectiveness
vs Will-o-wisp w/AC 19 and 22 hp, overkill is an issue, and the high AC means GWM=suck
vs Ghoul w/AC12 and 22 hp, overkill is an issue, but low AC helps GWM
vs Hill giant w/AC 13 and 105 hp, no real overkill problem here vs this big bag of hp, and low AC makes this a good target of GWM
Also, consider that many feats, powers and spells are more deadly in combinations. You have Advantage in your example, something very effective as a combo w/a lot of feats and powers. I think in the case of SS/archery its a bit more clearly low hanging min-max fruit, but GWM is a bit harder to make that argument for. I would argue there are hundreds of other equivalently deadly combos, both for players and DMs.