There are two issues of interest, at least to me 🙂 . the issue addressed specifically by this thread is the use of the AF assist beam from a speedlite. my testing shows, on my own 5D3, that this works very well. at essentially "pitch black" conditions, my 5D3 works beautifully with AF assist, where the 580EX ii paints a nice pattern of light on the subject,enabling the camera to lock focus instantly. within reasonable distances of course;
the other issue at work is the AF performance itself in low light, without AF assist from a speedlite. Canon specifies that the AF system is sensitive to EV -2, using the center AF point, f/2.8 and I assume a high contrast subject.
I performed an approximate test of this as well, by creating a lighting condition requring .5sec at f/2.8 and iso 6400 to obtain a good exposure (thats EV -2 if my calculations are correct.), and my 5D3 captured focus just fine (I even acheived focus at EV -3 but it wasn't dependable). it takes a while (sometimes a few seconds) but this is, in my opinion, reflective of the accuracy and retry-credence algorithms present in the camera. I have no problem taking that long to focus in that level of light because I'm usually not in a cave :-). At that level of light, you can hardly see the subject anyway, much less a moving one :-). keep in mind that the EV -2 sensitivity requires the center point.
in some ways, my tests were crude, i.e. I depended on the cameras own meter to roughly approximate EV -2. In other ways, my tests were severe, in that I was using a 100mm macro f/2.8 handheld. I thought it was impressive to see the AF system lock onto what it surely sees as a moving target with very little light. FYI I was using the default as-shipped focus mode.
I repeated the above tests with my 70-200 f/2.8 IS II, obtaining similar results. At approximately 15 feet subject distance and approximately EV +3, the AF assist even enabled focus on a very low contrast, dark subject (a treadmill belt, lol).
