You must be blind so I'll try to explain it again.
No one plays actually plays Crysis Warhead anymore. It's become a benchmark tool just like 3DMark, Vantage and Unigine which the 6970 loses dramatically in as well.
BC2, Mafia II, Just Cause 2, Lost Planet 2, Metro 2033, HAWX 2 and AvP are games people actually play.
The 6970 even loses to a 560 in single card comparisons... Only in 2-way do they become equal at 1080P and 1600P. 5760x1080 is a completely different story though.
You've clearly never played a FPS with an average of 44FPS before which means it's frequently dropping well below that. You must not be playing multiplayer or actually taking the game serious. You still can't even play with 8xAA with the 6970s so forcing that setting is completely pointless.
the 6970s finally become sort of playable yet are still frequently dropping below 60fps making gameplay stuttery which would lead to poor accuracy and higher deaths compared to 85fps which has plenty of headroom available.
At 5760x1080 it's the exact opposite. The 6000s have some serious scaling issues in half the games and when they don't, they still don't ever perform better than 570s even with the extra 720MB frame buffer:
^ A game where the 720MB less frame buffer finally affects performance yet still only loses by 1fps...
You don't even want me to post the synthetic benchmark results showing even worse scaling...
Let's take a look at the review you linked to.
A single 6970 losing to a stock 570? No way:
Uh-oh, what's this? Three 6970s equal to 3 570s at stock clocks(lol) from the site you linked to?
2 6970s @ 940 could already get 60FPS with 4xAA so 3 6970s with 8xAA only getting 68fps just goes to show how pointless stock reviews are. They could do much better if they were OCd to 940.
Do you even check the sites you link to before trying to make things up? This is yet another review further showing how the 6970 is weaker individually but due to its awesome scaling equals out to 570s in 2/3-way.
HH's 67FPS proves the 1280MB isn't limiting it in 5760x1080. Nvidia's extremely poor scaling is the only thing holding them back.
Here's another food for thought: when linking to review sites testing 2-way/3-way, most of them are testing with CPU bottlenecks in place by not overclocking to 4.5-5Ghz for the tests. Even this 1100T @ 4.2 with its lower IPC compared to a 2600K still has bottlenecks in most of the games at 3-way:
If a game is highly threaded, 4Ghz on a 1100T is equal to 4Ghz on a i7 9xx yet go look at the CPU clock speeds used in most of those 3-way reviews and see they're not even at 4Ghz. Most are at stock.
The HH review has a 980X at stock... While 6 cores is nice, most games have at least one thread that if bottlenecked affects the entire gameplay FPS. SC2 is one of the worst offenders of this and BC2 requires 4Ghz at least for its main thread; check task manager to see if any of your cores are at 100% to see it yourself.
Another win for ATI here: Fermi is notorious for requiring higher clock speeds to make use of their cards. 3Ghz just won't cut it.