In doing a few more tests I still have to say (even with the camera set to the lower 10 fps setting) AI-SERVO is nowhere near the advertised -2 EV. For instance when I set my camera to ONE-SHOT in 0 to -2 EV light the camera bangs focus perfectly. However, even set at the lower 10 fps in AI-SERVO under the same light condition, the camera fails to focus in that same light. It is certainly better than when it is set to 12 fps, but still cannot focus at 0, -1, or -2 EV light in AI-SERVO mode.
Can others confirm this? The short of it is that I don't really trust this camera to focus track in low light/contrast.
A big shame because this camera blows away anything I've ever used. It is just that good in every other respect. People say it's not that much different than the 5D3, but owning both I disagree. It is on a whole other planet from the 5D3. Don't get me wrong the 5D3 is a fantastic camera, but can't touch the 1Dx in nearly every area (including the AF). They share similar 61 AF system, but it isn't fully the same.
Really, the troubling low light/contrast AI-SERVO problem is really the only weakness I see with this camera. If it didn't affect the way I shoot I would just forget about it and move on.
It's not only you - reception hall - weddings and runways is where I am fearing it will fail.
With that said - I can confirm -2ev single shot af lock no real issues. Ai-Servo stops working a long time before that, however - it needs contrast to see the target, - 2ev is rather dark I at-least the scenario i created - I could barely see anything moving with my own eyes (the black af points don't help here either) let alone the camera being able to see any contrast at those light levels.
I owned a 2x5d3's for a short while and i didn't really put the unit through the paces as I did with the 1dx - yes its a whole other planet in terms of speed, there is no comparison between the two bodies. The single chip solution really lags the 5d3 (and tbh 5 fps is really more like 3 on the 5d3 with the same configuration i have my 1DX set at 12fps with no noticeable drop in fps). I the IQ between the two bodies esp at hi-iso is no comparison at least in my eyes.
From a technical standpoint the af points on the 1DX and 5D3 are very-very small. More suited to afpoint-to-target point to point hand off under normal light.
In comparison to nikons offering where the center af cluster are enlarged af points - low light -2ev targeting is much easier (only thing i had to compare was a d800) where the nikons offering locked 8/10 attempts where the 1DX locked 5/10 attempts. At -1ev the the canon was much faster and hit 10/10 (and about 50% faster to lock then the d800)
With that's said useable single shot on the 1DX is -1.7-1.8ev D800 really drops off around -1-1.2ev (for the time it takes to lock reasons)
But there was absolutely no difference in 3d tracking vs single shot on the nikon where the 1DX drops off a long time before that.
I can't re-run the same tests as I dont have access to a d800 anymore but I would be interested to see if the same test run in Ai-Servo L with return the same results.