DXOMark Measurements differences between screen and print

[vscd]

I've got a question to sensorbased measurements at DXOMark. As dpreview seems to have quite good contact to DXOMark I would like to ask about the differences between the the "screen" and "print" switch in the measurements.

If I switch to the "screen"-tab the results seem to be normalized to 8MP, which roughly is the resolution I work with. Even most clients get their pictures around 8 to 12 MP.

So do I understand it right that my old Canon 5D is better than even the new Nikon D810 (except in DR in low ISOModes)?

http://www.dxomark.com/Cameras/Comp...Nikon-D810-versus-Canon-EOS-5D___1008_963_176
I sometimes really got this feeling, especially in Color sensitivity, but now there seems to be a valid chart for this. I know the differences between the bodies apart from the sensor, but anyhow, pixelpitch and Low Resolution is still better than newer sensortech and high resolution?

Or do I understand the results in a wrong way?

Edit: For Example: http://tf.weimarnetz.de/downloads/example.jpg

 

[Steen Bay]

I've got a question to sensorbased measurements at DXOMark. As dpreview seems to have quite good contact to DXOMark I would like to ask about the differences between the the "screen" and "print" switch in the measurements.

If I switch to the "screen"-tab the results seem to be normalized to 8MP, which roughly is the resolution I work with. Even most clients get their pictures around 8 to 12 MP.

No, "screen" is the 'pixel quality' when viewing the full resolution images at 100%, "print" is the 'pixel quality' when normalized (downsampled) to 8mp, which makes it an apples to apples comparison.

P.S. - Hold the mouse over "screen" or "print" to see an explanation.

So do I understand it right that my old Canon 5D is better than even the new Nikon D810 (except in DR in low ISOModes)?

http://www.dxomark.com/Cameras/Comp...Nikon-D810-versus-Canon-EOS-5D___1008_963_176
I sometimes really got this feeling, especially in Color sensitivity, but now there seems to be a valid chart for this. I know the differences between the bodies apart from the sensor, but anyhow, pixelpitch and Low Resolution is still better than newer sensortech and high resolution?

Or do I understand the results in a wrong way?

Edit: For Example: http://tf.weimarnetz.de/downloads/example.jpg

 

[vscd]

I read the explanation and it means to me if I work with resolutions not more than 8-12MP any old cam with less resolution gives better results for me.

The only exception is the marvelous A7S with 12 MPixel and new technology, too.

 

[Lee Saxon]

No, that's not what that means.

Take a look at the SNR chart in "screen" mode. That shows all three cameras at full resolution / 100%. That basically shows you the quality of each individual pixel on the camera / the "per pixel" quality.

Now, let's say you decided to make ~12"x8" test prints (the largest the 5D can make at optimal [360dpi] resolution).

The way to test per-pixel quality ("screen mode") would be to crop the middle 13 megapixels of the 5Ds and D810 images. That way, each camera would have the same number of pixels represented in their respective prints.

In that scenario, then "per pixel" quality wins out and yes the 5Ds print is slightly noisier than the D810 print which is slightly noisier than the 5D print.

HOWEVER. Almost no one actually crops that heavily on a regular basis. And if they do they should just save money and weight by using a m4/3 camera.

No, what's almost always going to happen is that you'd downsample the 5Ds or D810 to the size you want (in this case ~13mp). In that situation "per pixel" quality matters drastically less because you're averaging a lot of pixels together to get a few better ones. This is called "oversampling."

"Print mode" simulates oversampling, and you'll notice that the rankings are quite different.

 

[vscd]

Thanks, interesting point. That's maybe the missing link for me. Never saw it that way.

But if I just need the 12MP it should be fine to still use the 5D. I'm still surprised how good the sensor is (except of CR, which I can handle).