DxO Mark explained

[dilbert]

http://www.luminous-landscape.com/essays/dxomark_sensor_for_benchmarking_cameras2.shtml

... includes specific discussion relating to Canon's "flatlining" in sensor performance and that the 5D3 bests the D800 above ISO 3200.

[Zlatko]

We learn that the overall Camera Sensor score is biased toward single shot HDR at low ISO settings — a capability we never had in the past and which we may infrequently need (depending on what we photograph).

We also learn that the "Portrait" score is misnamed.

[meli]

Great article, which will be promptly discarded unread by some users who want to believe their camera is the best and no amount of scientific data will change that, or some other blondes that consider punctuation errors as a serious hint against their data reliability 

[Mt Spokane Photography]

Excellent write up, and it matches my experience with my D800 and 5D MK III sensors.  The D800 sensor is supurb at low ISO settinngs below 400, while my 5D MK III pulls away at high ISO settings.
Since I do mostly extreme low light photography, I was disappointed with the D800 performance in low light, but amazed when I used ISO 100.
The thing that is not touched on, are the aspects of a camera system.
Reliability
Service (time and quality for repairs)
Lens quality and available focal lengths that match sensor resolution.
There are other personal preference things like ergonomics, ease of use which tend to vary with user background.
He did touch on the issues with editing of large raw files, but a lot of people do not know exactly how to quantify this.  I found it taking multiple times to render a low light D800 image compared to a 5D MK II image, and running NR or other computation intensive processes was very slow when you were trying to process 500 or more images.

[Mt Spokane Photography]

Excellent write up, and it matches my experience with my D800 and 5D MK III sensors.  The D800 sensor is supurb at low ISO settinngs below 400, while my 5D MK III pulls away at high ISO settings.
Since I do mostly extreme low light photography, I was disappointed with the D800 performance in low light, but amazed when I used ISO 100.
The thing that is not touched on, are the aspects of a camera system.
Reliability
Service (time and quality for repairs)
Lens quality and available focal lengths that match sensor resolution.
There are other personal preference things like ergonomics, ease of use which tend to vary with user background.
He did touch on the issues with editing of large raw files, but a lot of people do not know exactly how to quantify this.  I found it taking multiple times to render a low light D800 image compared to a 5D MK II image, and running NR or other computation intensive processes was very slow when you were trying to process 500 or more images.

 i have them, after 12800iso the 5dmk3 is better regarding signal/noise. now the 800 is not aimed to be a high iso camera but does it rather well

At the time I ought it last May, I fell for the supurb high ISO pitch that was going around.  I found that at ISO 800 and above, noise started to show up.  I took a large number of images with it at ISO 12800 in very low light.  The reduced High ISO DR is a problem for all cameras at very high ISO, but I was disappointed with the D800.
I also was disappointed in my Nikon 24-70mm f/2.8 G, it may have just been a problem lens, but the CA was horrible.  I like to use live view and tether my cameras, this turned out to be another Nikon weak point, the screen was very slow to update makiing manual focus difficult. 
All those things and a lot more were overlooked in the hype.  I see it as a wonderful camera for landscape shooters, but nothing special for high ISO use.
 

[Drizzt321]

the d800 is equal to 5dmk3  up to 12800iso, i have them , i have tested them and compared them step by step and Im talking raw files.

this is 5dmk3 together with 70-200/4 and Nikon d800 with the new zoom 70-200/4  6400iso, my purpose was to se if Nikons new zoom is good as Canons.
Nikon file size down sampled to Canons 5dmk3 and  latest Camera Raw
canon to the left

Not saying your method isn't valid for you, but remember as you sample downward, you remove more and more of the noise that appears. So if you regularly use the D800 files at the 5d3 size resolution, you may see similar or better performance for some of the higher ISOs, while if you often use the full resolution, you'll be more likely to see the noise in the D800. Try down sampling both to 18, or 12, or some other size and see what happens to the noise in both. For example, when I export for the web (1200px long edge), often much of the noise disappears, unless it was truly horrible to begin with.

I think this is the real power of the newer generation(s) of high megapixel-high ISO sensors. The ability to get pretty good files straight out of the camera at full resolution, and be able to down-sample to a smaller resolution and get stunning results in lighting that would previously be considered next to impossible to shoot in.

[NormanBates]

Comparing noise at different resolutions makes no sense whatsoever. My one pixel camera has no noise at all, even at very high ISO; will you buy it from me?

[heptagon]

Comparing noise at different resolutions makes no sense whatsoever. My one pixel camera has no noise at all, even at very high ISO; will you buy it from me?

Comparing the noise at the resolution you need makes perfectly sense. You need a 800 by 600 pixels image? Take any camera you get, crop the image and then scale the image to 800 by 600. Then you can really compare which camera/lens combination delivers you the best results for that specific resolution.

[LetTheRightLensIn]

We learn that the overall Camera Sensor score is biased toward single shot HDR at low ISO settings — a capability we never had in the past and which we may infrequently need (depending on what we photograph).

We also learn that the "Portrait" score is misnamed.

As if people ever used to run around shooting ISO 25,600 or even 6400....
B&W film DR was closer to high DR low ISO than any film was to what people do with high iso today.

Anyway, the thing is that is why they don't just give an overall score but individual plot so you can focus on what matters to you, which what almost every DxO proponent has said to do all along.

[LetTheRightLensIn]

Excellent write up, and it matches my experience with my D800 and 5D MK III sensors.  The D800 sensor is supurb at low ISO settinngs below 400, while my 5D MK III pulls away at high ISO settings.
Since I do mostly extreme low light photography, I was disappointed with the D800 performance in low light, but amazed when I used ISO 100.
The thing that is not touched on, are the aspects of a camera system.
Reliability
Service (time and quality for repairs)
Lens quality and available focal lengths that match sensor resolution.
There are other personal preference things like ergonomics, ease of use which tend to vary with user background.
He did touch on the issues with editing of large raw files, but a lot of people do not know exactly how to quantify this.  I found it taking multiple times to render a low light D800 image compared to a 5D MK II image, and running NR or other computation intensive processes was very slow when you were trying to process 500 or more images.

 i have them, after 12800iso the 5dmk3 is better regarding signal/noise. now the 800 is not aimed to be a high iso camera but does it rather well

At the time I ought it last May, I fell for the supurb high ISO pitch that was going around.  I found that at ISO 800 and above, noise started to show up.  I took a large number of images with it at ISO 12800 in very low light.  The reduced High ISO DR is a problem for all cameras at very high ISO, but I was disappointed with the D800.
I also was disappointed in my Nikon 24-70mm f/2.8 G, it may have just been a problem lens, but the CA was horrible.  I like to use live view and tether my cameras, this turned out to be another Nikon weak point, the screen was very slow to update makiing manual focus difficult. 
All those things and a lot more were overlooked in the hype.  I see it as a wonderful camera for landscape shooters, but nothing special for high ISO use.

Compared properly, few have found the 5D3 to do any better until you get neareing ISO6400 compared to the D800 and the difference is not as large as the difference at low ISO. What matter more though depends on the individual. If you want to talk fps at FF-size or left AF points working or liveview quality and such it may be a different matter.

Anyway hopefully this finally puts to rest all the talk from jrista about how the Print plots are the ones to ignore and that the Screen plots are the ones to use when you want to compare two cameras when the reality is the opposite, Print to compare two cameras, Screen to see what a camera does at full resolution.