looking at some AF tests on some forums, early tests, rather brief, etc. etc. keep that in mind, hint at a really humungous improvement in AF. In one test a guy focused a fast lens, wide open, barely any DOF on his little girl's eye and the center point 5D2 was dead on sometimes and somewhat off sometimes only truly nailing focus about 50% of the time, the 5D3 was dead on pretty much every frame, considerably better performance; using the left point, the 5D2 was off on by far most frames, looked like 80%+ off, while the 5D3 appeared to be just about dead on on every single frame. 
So the AF earlier results, at least for non-sports tracking, are showing some very encouraging signs. I also saw one guy post a perfectly focused shot of a backlit shopkeeper with tons of brigher and more contrasty and detailed stuff behind him, a test that canon af has often failed on, certainly below 1 series, so that was looking great too.
OTOH, one of the guys posted a high ISO comparison and yeah the 5D3 was better, but wow it looked closer to only 1/3rd stop better.
Granted using ACR is NOT the best way to test and maybe at the super duper high iso above ISO6400 the story would be different. But if this result held up then we have basically the same sensor as the 5D2 with just a few minor tweaks while nikon put much process sensors in the D4 and D800, even the D4 a non-exmor, managed to improve 1 stop low ISO DR over a few years, but Canon not a single bit :'(. It seems Canon is investing less in sensor fab while charging a lot more for their equipment now. :-\
So we are facing a new situation, we are like the Nikon of old, body performance may be fantastic but the sensor kinda pales in comparison (although maybe at ISO1600+ it will be top notch current so it may only be behind on one end of the scale, IF it does get to D3s-type leves on the top end that is certainly a nice improvement, no doubt).
Hopefully the more careful tests of DxO will paint a better picture of the High ISO tale though. And we will at least get high ISO up to D3s,D4 standards. This guy's test looked a bit worrisome, looked questionable to even match the D800 at high ISO, but other tests make it seem quite possible. ANd it's all but been established that the D800 will be noticeable better at ISO 100 for scenes having large dynamic range (they really are not that hard to come across, in part, people were so used to thinking them impossible, they don't even see them anymore in their photographic mind's eye and automatically right them off without even thinking about it).
But the early AF tests are looking VERY, VERY good so far. The body performance seems great this time.
ADD: I've since seen reports from a few people who shot a game or two and they said it did better than their 1 series body when it came to AF.
ADD: The video is apparently blurry and not real 1920x1080 again, getting blasted on the video pro blogs apparently, if they couldn't do 3x3 smoothtly since those blocks are too large compared to the AA filter without having to blur down res to stop aliasing then why not offer a second, cropped mode, using 2x2 blocks like the C300? That may have made the video on this totally killer and been better for the wildlife videographers just in general.
So the AF earlier results, at least for non-sports tracking, are showing some very encouraging signs. I also saw one guy post a perfectly focused shot of a backlit shopkeeper with tons of brigher and more contrasty and detailed stuff behind him, a test that canon af has often failed on, certainly below 1 series, so that was looking great too.
OTOH, one of the guys posted a high ISO comparison and yeah the 5D3 was better, but wow it looked closer to only 1/3rd stop better.
Granted using ACR is NOT the best way to test and maybe at the super duper high iso above ISO6400 the story would be different. But if this result held up then we have basically the same sensor as the 5D2 with just a few minor tweaks while nikon put much process sensors in the D4 and D800, even the D4 a non-exmor, managed to improve 1 stop low ISO DR over a few years, but Canon not a single bit :'(. It seems Canon is investing less in sensor fab while charging a lot more for their equipment now. :-\So we are facing a new situation, we are like the Nikon of old, body performance may be fantastic but the sensor kinda pales in comparison (although maybe at ISO1600+ it will be top notch current so it may only be behind on one end of the scale, IF it does get to D3s-type leves on the top end that is certainly a nice improvement, no doubt).
Hopefully the more careful tests of DxO will paint a better picture of the High ISO tale though. And we will at least get high ISO up to D3s,D4 standards. This guy's test looked a bit worrisome, looked questionable to even match the D800 at high ISO, but other tests make it seem quite possible. ANd it's all but been established that the D800 will be noticeable better at ISO 100 for scenes having large dynamic range (they really are not that hard to come across, in part, people were so used to thinking them impossible, they don't even see them anymore in their photographic mind's eye and automatically right them off without even thinking about it).
But the early AF tests are looking VERY, VERY good so far. The body performance seems great this time.
ADD: I've since seen reports from a few people who shot a game or two and they said it did better than their 1 series body when it came to AF.
ADD: The video is apparently blurry and not real 1920x1080 again, getting blasted on the video pro blogs apparently, if they couldn't do 3x3 smoothtly since those blocks are too large compared to the AA filter without having to blur down res to stop aliasing then why not offer a second, cropped mode, using 2x2 blocks like the C300? That may have made the video on this totally killer and been better for the wildlife videographers just in general.
