Well keep in mind, the optimizations will keep coming and so will the performance. Free performance is nothing to be negative about. I don't quite understand peoples views about wanting Mantle to fail. I guess if i paid $200 more for a video card that got a few fps more, i would be pissed. But, shame on them, for making a stupid decision anyhow.. E-peens get you no where.
My concern isn't really related to anyone, myself or some random person on the internet with a different car, getting better performance. I'm completely all for that.
What I am concerned about is the implication that benefit in performance might lead to. I couldn't really care less what frame gains might be realized with Battlefield 4. My concern is the case of Oxide. I like RPG and RTS games. I like Stardock and similar games. I appreciate the mission statement that the engine used in/Oxide engine was programmed from the ground up to work better with 64 bit high memory systems, because RPGs and RTS titles need different things than a military FPS. I get all of that.
But what I am potentially worried about is that, if Mantle gets enough traction games like these could become just as much about showcasing the strengths in that API and degrading performance on DX as about making great games. Traditionally, when one vendor or another comes out with a feature or an optimization, it's in an engine or for a game that's still DX compliant, so if you aren't in the right camp for that feature or optimization, hey...no biggie, you can still play the game. NVIDIA has TXAA, AMD has MLAA, and so on and so on. These features are nice for people who like them, but not necessary, and aren't going to cripple a game for the other camp of people because they are on as a mandatory.
Mantle has a chance to change that. Having tried to run the Oxide demo here on the machine I am at currently, I'm kind of concerned. Turning that down as low as it can go, upscaling, etc, nets me horrible slide show performance. Granted this is on one of my work workstations, with a mere mid range i5 and a 6770. But, on this same machine, I can run Shogun, Civ 5, and a number of other, potentially unit heavy games, albeit at machine reasonable settings. I'm concerned that that won't be an option in the near future. Effectively, I'd be forced to buy a very specific combination of hardware in order to play certain titles even at mediocre settings, because if you don't your DX/OpenGL only hardware will get hammered into submission by overaggressive draw calls and effects like extremely computationally heavy motion blur, effects that you can't turn off, unit numbers that you can't turn down.
I'm also not a fan of that logic getting extended even further. Mantle looks like it's going well, so NVIDIA gets in on the act, and underwrites a different set of studios, and gets another API. Then, heck, Intel, feeling a little CPU heat, decides to use a partnership with Microsoft to get MMX like calls written into compute in DX 12, so that AMD chips run like junk, AMD responds by making Mantle AMD CPU specific as well....etc.
So while more detail could be gone into, no, I'm not in any way upset that some people are going to get better performance out of their hardware. Great. I'm slightly bummed that my 6950 crossfire system is out of that running, and may never get frame pacing benefits, but hey...whatever, I've moved on. Ideally, I would love for Mantle to create enough stir to satisfy people that had a desparate need to be justified in their hardware choice, and also to rouse the lazy giants at the OpenGL and DX standards bodies. Enough to get some real, more widely applicable optimizations and reworking of code paths done. The majority of what developers have said about the Mantle API seems to be related to good coding practices, engine optimization, and tools. Those should in theory be applicable to any hardware given the right adoption. And maybe they will be, if AMD would stop being cagey about what the 'minimum requirements' for Mantle adoption are.
I hope that wasn't too long winded, but seeing as real performance analysis is missing along with the drivers, it seemed like a good place to interject some perspective about
why I might be interested in the outcome of the roll out.