We got a couple of Tamron 24-70 f/2.8 VC in Canon mount today and I was able to compare them, very briefly, with the Canon 24-70 f/2.8 L.
The Tamron is not quite the range of the Canon on the long end, clearly a few mm shorter than Canon (which is clearly a couple of mm short of 70mm). For anyone who doesn't know, manufacturers "round" up or down to give the numbers they actually put on the zoom. I can't see where being a bit shorter is going to affect anyone very much.
Autofocus is reasonably fast, although the Canon may be a bit quicker. Nothing dramatic like the Tamron 70-300 (which was horribly slow), though. Again, I can't see this making a ton of difference to anyone.
Imatest results were clearly in favor of the Tamron, though, at both the long and short ends. I was able to run 3 copies of the Tamron on 2 cameras and results were as consistent as we'd like them to be.
These are MTF 50 results (3 copies of each lens tested, best results of each copy averaged, variation +/- 2.5%) measured in line pairs / Image height on 5D Mk II test cameras from unsharpened raw files. Ctr = the Center Point, Avg = a weighted average at 13 points on the lens (center, 4 mid 1/3, 4 corners, top and bottom) The Tamron is clearly better on the wide end, particularly in the average number across the entire lens. At 70mm the centers are about the same (you couldn't see that difference, even in reasonably sized prints), but the average across the lens is again better with the Tammie
The Lens sharpens up even further at f/4, and a bit more, particularly in the corners at f/5.6 as shown in the graph below.
So what does this all mean? It means the lens has potential to be good. We'll have to wait until some real world results and reputable reviews come in to get some real-world initial impressions.