6000 series was definitely ahead of NVIDIA when it comes to power consumption. They were about equal in terms of price

erformance, but NVIDIA had the big 580 powerhouse that AMD couldn't compete with at the high end.
The discussion has changed from FPS metrics to frame latency, as some pretty interesting information was uncovered by the Tech Report. tl;dr, the way that FPS information is polled, it tends to hide a ton of data where things are going wrong, and has been found to be nearly useless when it comes to looking at performance.
AMD was found to be screwing up the memory interface which led to very stuttery performance, despite high FPS numbers. The reason why this existed is because their internal team had never even tested for frame latency, sticking to the same metrics that most review websites were using instead.
They've since addressed the problem, admitted fault, reasons for why they screwed up, and came out with a patch in a matter of weeks. Pretty fucking rad right there. It's supposed to continue to improve, as it's not exactly perfect yet.
Where NVIDIA kind of dropped the ball, so to speak, with Kepler is that they've basically been releasing what were initially tagged for mid-low range cards as the mid-high range cards. Because of this, a lot of the mid range ones have some pretty severe memory bandwidth issues, and they end up choking pretty hard when a game uses a high amount of VRAM, or you crank up the AA/Resolution. Basically everything from the 660Ti and below is only suited well to 1080P performance with low AA.
To top it off, NVIDIA's cards don't scale really well in SLI due to the limited memory bandwidth. With this in mind, the really high end setups for benchmarking/performance are pretty much limited to AMD right now, unless the person is just totally sold on NVIDIA as part of their identity. I'm talking about triple display type setups in this case.