Alright, time for some concrete evidence. Here is a 5-frame bracketed sequence I took yesterday of a sunflower field at sunset:
The shots were bracketed 2 stops apart. I was shooting directly into the sun, as you can see. Did I really want to? Well, kind of. More importantly, I HAD to. The sunflowers face east, towards the rising sun. To actually have sunflowers in my photos, I had to shoot directly into the sun.
Here are the first, middle, and last images:
The first was exposed for the sun. The last was exposed for the foreground. Now, since everyone want's a fair comparison, I'm using the middle image from the 5-frame sequence as the "most fair" towards my 5D III. Here it is after processing it to extract the most detail possible:
This is a +3 stop exposure pull (lift), -100 highlights, -100 whites, +60 shadows. As you can see, the sun is blown, and it has some posterization around it. Here is a closeup of the sun:
Here is a closeup of the noise in the foreground:
That's not good noise. That's nasty noise. It's banded, red-shifted, and it is already getting blotchy. I also took the image exposed for the sun, and did a +5 stop exposure pull, -100 highlights, -100 whites, +75 shadows, +10 blacks:
In this shot, the sun is MUCH better, although it's still hot and overexposed in the end. However, the foreground...the foreground is absolutely atrocious. The 5D III can MAYBE handle a +3 stop pull, but it definitely can not handle a +5 stop pull. If one were to listen to dtaylor about Photographic DR, one would believe that Canon cameras are only within a third-stop of Exmor sensors...however, who on earth would consider the above image acceptable? Anyone? I mean, be truly honest here.
Here is a close up comparison of the noise, color fidelity, and detail of the +5 pull, +3 pull, +3 pull denoised/debanded, and the last shot that was exposed for the foreground:
The +5 stop pull, as well as the +3 stop pull, both suffer from the fact that a lot of the detail is buried within the noise floor. I worked the +3 stop pull for about 10 minutes. I could probably reduce noise further, however I was already losing detail (look at the edges of some of the leaves deeper in the shadows of the +3 denoised...compare them with the +3 non-denoised and the full exposure...LOT of softening!!) I did not even bother to denoise the +5 stop pull...that is well beyond the capabilities of ANY Canon camera. Since most computer screens are 8-bit, a 3-stop pull is necessary to fully realize the 11 stops worth of DR in a Canon RAW, without lifting the read noise as well. Well, based on the samples here, even a 3-stop pull is kind of pushing it...the banding and color blotchiness is all read noise...photon shot noise exhibits as clean, random noise (of which there is definitely plenty, but it's mixed with read noise as well.) A 5-stop pull would be necessary to fully realize the 13 stops worth of DR that a D800 has. I don't think anyone honestly denies that Sony Exmor sensors have more DR. Aside from dtaylor's Photographic DR, the standard definition of Engineering DR, the ratio between the clipping point and the RMS of read noise (the read noise floor) indicates that Exmor sensors have a two-stop lead on Canon sensors. So...with a D800...I could have made this photograph in a single shot. Directly into the sun, underexposed by five stops, then lifted five stops.
Finally, here is the HDR merge:
I had problems with this. It was a 5-frame sequence, separated by 2 stops. I ran into problems with posterization around the sun, and even after the HDR merge, I could not fully realize the sun. I was able to recover the sky and all the other detail, but still not the sun. Given the nature of the posterization, I figure I would have needed at least a 9-frame sequence separated by 1 stop or less to produce a fine enough grade in the highlights around the sun to avoid posterization. Some HDR wizards would have probably shot 15 frames. Either way....it's more work. Getting a 15-frame sequence for HDR that does not run into problems with motion in the scene...the sunflowers moving in the wind, the clouds moving, even the sun moving as it sets, is more difficult.
And, I could have gotten it in a single shot with a D800. Here is a single-shot sunflowers into the sunset shot with a D800 on 500px:
http://500px.com/photo/40685186/sunset-flowers-)-by-kenji-yamamura