As with the Nikkor, the Canon presented itself with two personalities: on the 40D it is soft most everywhere at most all settings, but on the 5D it is sharp enough everywhere at most all settings. On the 5D it is a very well balanced design that handles superbly, but on the 40D it just fails to deliver in my opinion for the price. The question is whether this performance on FX format would extend to a 1Ds Mk III as well, where the performance might resemble more that of the 40D due to the much higher pixel density.
To put the Canon’s sensitivity to pixel density into perspective, on the D300 the Nikkor performs as well or better wide open (f/2.8) than the Canon does on the 40D at its peak aperture (f/5.6 - f/8). Note that the D300 has 20% more pixels than the 40D! The 1Ds Mk III has 75% more pixels than a 5D… so the performance question hasn’t been fully answered by the Dpreview test.
Here follow some DPReview charts, here's the decoder ring: blue=excellent, green=good, yellow=poor, red=terrible.
DX at 70mm, f/2.8. Left: Canon. Right: Nikkor. Oops Canon.
On “full-frame” the Canon performs acceptably wide-open, but really comes into its own starting at f/4. The Nikkor doesn’t really benefit from stepping down as much. In comparison the Nikkor’s performance wide-open is superior to the Canon’s, except in the corners where it is never good no matter what aperture you’re using. The Canon extends its performance all the way to the corners, and possibly all the way to the edge of the image circle.
FX at 135mm, f/2.8. Left: Canon. Right: Nikkor. Oops Nikon.
The best charts
The Canon is simply sublime here, but there's nothing to be ashamed of on the Nikkor side.
Left: Canon 100mm f/4 on FX. Right: Nikkor 70mm f/4 on DX.
The worst charts
Neither are acceptable but I personally prefer the Nikkor. I can avoid putting details in the corners, but I can't use a all-round soft lens.
Left: Canon 200mm f/2.8 on DX. Right: Nikkor 200mm f/2.8 on FX.
I can't entirely agree with Dpreview's ratings here. Granted, the Nikkor's performance in the corner is bad... but the Canon simply isn't very sharp. Is that better? If a measly 10MP sensor on DX format can put it down for the count, then I think its days are counted even on FX format.
We're not likely to see many more 12MP sensors on FX format, so the 5D isn't the most representative body to test this lens' long-term value.