They've been working on the dx version even longer and frostbite in general for years, hasn't stopped them making a complete mess of it.
If dice cant figure out stability on an api that's been around years I'm sceptical they'll crack an api that's only months old at the first attempt.
While this is true to a degree, and keep in mind I'm not saying Mantle is the end of all bugs, it's possible that Mantle is much simpler to get working sans bugs, because DX is a massive pile of complex turd, where drivers, dx, updates to all, and the game have to do a bunch of work arounds to get them all working correctly.
Mantle driver is simple without doing an awful lot, it's down to the game. When only trying to solve bugs that can come from one source it IS easier to get right. That doesn't mean DICE will.
The timing of Mantle releases, and the quality has precisely NOTHING to do with how good or useful Mantle is.
Whether a game is good, or crap, buggy, or flawless, 30% faster is 30% faster, 3% faster is 3% faster, that's it. Expecting huge changes and bug fixes because of an API, and frankly misunderstanding and misstating things isn't helping.
Dice have NOT worked with mantle for ages and it wasn't a part of the original engine. Dice got a release quality build 2 months before their talks at APU 13, as did Oxide, Dice and other dev's worked with AMD, but this is mostly on the spec of it. Any software design starts with going to the people using the software and asking them what they want in it, then they go away and figure out how to implement it based on these design features. Many things change, but AMD wrote and made Mantle and Dice got it now 2.5-3 months ago now.
Look at a console on release vs a console 8 years later, night and day, they use 100% power available to them in both cases. This is hardware experience, software experience, api experience, and generally people getting better at their jobs over time. As in any other job, you learn things and get better at doing whatever it is you do. Mantle is new, and it's pretty much the only ever attempt at a low level API on an open platform with chips this complex. Glide may have conceptually done the same thing, low level API... but chips then vs today, night and day, completely night and day. That is just the chips, games are exponentially more complex now as well as the hardware.
The consoles, while low level, are a very set platform with no scaling to think about at all, no different hardware, nothing. Expecting perfection from Mantle out of the gate when no other API has ever done such a thing, no simpler API has ever done it, and everything is more complex now... is just ridiculous.
I'm not saying Mantle will be terrible at first, but expecting the impossible based on no history of this being possible is nothing short of ridiculous.
The only thing that matters, is not launch dates, nor quality of games, but if dev's like using it, and if it enables the dev's to do more with the same hardware, if yes to one or both of those, win, if neither of those, lose.