Canon 1D-X: "Full" 2-Stops ISO Improvment...only in JPEG??

[traveller]

A few points:

1. According to DXO Mark, the D3s reaches its "acceptable" signal to noise value of 30dB at 3253 ISO (whilst keeping dynamic range at >9 EV and colour depth at 18bits), whereas the 5D MkII achieves this at 1815 ISO. So 3253/1815 = ~1.79; i.e. the D3s is 79% better than the 5D MkII, which is basically just over 3/4 of a stop.

2. Nikon is implying that the D4 manages to achieve the same high ISO performance as the D3s, but with a higher pixel count. Thus, the 1D X would need to be between 3/4 and 1 stop better than the 5D MkII to match the D4.

3. I've got a suspicion that the D3s 'cooks' its raw files at high ISO, based upon the SNR 18% graph from DXO Mark having a strange kink in it after ~ISO 9787 (also note the performance of the D3).

Anyone want to check these calculations?

 

[cpsico]

The exact same story came out when the 1D MK IV was announced. Chuck Westfall said the 1D MK IV was 1-1/2 to 2 stops better than the 1D MK III in JPEG, but about 1/2 stop better shooting raw. It was due to the more sophisticated in camera processing that could be done using the Digic IV processor. Processing has improved, and Digic V is more powerful than Digic IV. Jpegs will improve over the 1DS MK III and 1D MK IV.

The question I have, is how much improvement in raw over my 5D MK II? 1/2 stop or 1 stop? If I can get clean raw images at ISO 12800, thats about 1/2 stop or slightly more than my 5D MK II, but if they are clean at ISO 25,600 then there is a significant advance, which I somehow doubt.

When I bought my 5D MK II in 2008, using LR 2 allowed me to shoot at ISO 3200, but results were noisy. Now, with LR 4, I shoot ISO 6400 images and with no NR at all, they are remarkably clean. Processing software has improved noticibly in the last three years, so I've gained about 2 stops when using raw, just by spending a few dollars on software upgrades.

1/2 a stop in raw with a 60 percent increase in resolution is amazing if you ask me, of course it goes to show if canon would pull there head out of there hind part they could build a a camera with amazing low light image quality, but I am not sure the average consumer would buy a camera with amazing image quality over a high resolution camera.

 

[JR]

A few points:

1. According to DXO Mark, the D3s reaches its "acceptable" signal to noise value of 30dB at 3253 ISO (whilst keeping dynamic range at >9 EV and colour depth at 18bits), whereas the 5D MkII achieves this at 1815 ISO. So 3253/1815 = ~1.79; i.e. the D3s is 79% better than the 5D MkII, which is basically just over 3/4 of a stop.

The Nikon D3s being only 3/4 of a stop better then the 5DmkII seems low. If your calculation are right, the hill the climb is much less then I thought...but I recall seeing some very high ISO sample from the D3s that just destroy the 5DmkII at 12800 and above. Not sure how these charts works, but something is weird here...

 

[traveller]

A few points:

1. According to DXO Mark, the D3s reaches its "acceptable" signal to noise value of 30dB at 3253 ISO (whilst keeping dynamic range at >9 EV and colour depth at 18bits), whereas the 5D MkII achieves this at 1815 ISO. So 3253/1815 = ~1.79; i.e. the D3s is 79% better than the 5D MkII, which is basically just over 3/4 of a stop.

The Nikon D3s being only 3/4 of a stop better then the 5DmkII seems low. If your calculation are right, the hill the climb is much less then I thought...but I recall seeing some very high ISO sample from the D3s that just destroy the 5DmkII at 12800 and above. Not sure how these charts works, but something is weird here...

Yes, there is something weird here and I think that it relates to my last point (3). If you look at the SNR 18% graph, the D3s line is higher than but parallel to the 5D MkII's line until at above ISO 6400 the D3s suddenly starts to pull away (from the D3 as well). Either the sensor's drop off in performance suddenly gets better at this point, or Nikon is starting to apply some noise reduction to their raw files.

 

[Viggo]

I would take clean high iso to get my shutterspeed fast enough over twice the mp any day, and that is just common sense;

If your image is fantastically high res, you will see every bit of that juuuust to long shutter speed you used, and you'll have a useless image. I can tell you, going from mk3 to mk4 with the same size sensor but 60% increase in res, I had to double my shutterspeed, giving me the exact same noise under the same exact light. Now, the pictures that did stick with the mk4 had much better detail, but I had a lot of shots I kept, but can't use for anything due to (unexpected) motion-blur.

The 1d X will be the ultimate camera for me, no matter what I am shooting, and I can shoot things I could never dream of with the mk4. After trying this new AF, I had chills for the first time using a camera. It is just astonishing...

But if you do landscape, why would you buy a Worldclass AF-system and 12 fps? Buy a 5d2 for your landscapes, very little will surpass it.

+1

The high MP bodies do need a tripod, good IS lens, or a faster shutter speed to actually realize the increased resolution. Even Canon mentioned this in one of their articles about shooting with the new high mp bodies. Of course, you can get lucky and hit on one of 8 or 10 at lower shutter speeds.

A higher ISO rating at existing MP lets you keep the same shutter speeds you are using and get better resolution, or it lets you increase shutter speeds if your current one is too low.

I'm expecting to hear the Nikon users scream as they adjust to a 36 mp sensor and find they need to double or triple shutter speeds to get the high resolution.

Excellent points, both! I think you guys hailed the consequence of higher resolution on the head, at least for photography where you can't or don't necessarily use a tripod.

Here's the article, or AF-guide, check out page 34 and 35. I tell you, I can't be more excited that Canon kept it at 18 giving me 21% larger pixels, and therefore (I don't know the real math of this) giving me 21% longer shutterspeed with sharp images.

http://www.canon.co.uk/Images/EOS%201D%20MK%20IV%20AF%20guide_tcm14-721275.pdf

 

[elflord]

The Nikon D3s being only 3/4 of a stop better then the 5DmkII seems low. If your calculation are right, the hill the climb is much less then I thought...but I recall seeing some very high ISO sample from the D3s that just destroy the 5DmkII at 12800 and above. Not sure how these charts works, but something is weird here...

one thing to keep in mind is that this test only shows how well the two cameras do at 3200 and 1800 ISO. DxOMark tests how high you can crank the ISO and maintain a certain level of IQ (I think defined in terms of dynamic range and noise)

It doesn't say anything about how well they hold up above that (how much they lose if you turn ISO higher), and there's no reason the curves should be parallel. One example is that medium format cameras do very well at low ISO but not as well at high ISO as full frames.

 

[tt]

Less than 2 weeks before we find out from the embargo lifted 1DX photos being allowed on Feb 7 if the rumor's right

 

[cfargo]

canon will get f£$ked by interviews like this one, well all is dommed now, good luck guys im jumping to the Nikon boat.

Good luck getting your Nikon serviced locally.

 

[elflord]

A few points:

1. According to DXO Mark, the D3s reaches its "acceptable" signal to noise value of 30dB at 3253 ISO (whilst keeping dynamic range at >9 EV and colour depth at 18bits), whereas the 5D MkII achieves this at 1815 ISO. So 3253/1815 = ~1.79; i.e. the D3s is 79% better than the 5D MkII, which is basically just over 3/4 of a stop.

2. Nikon is implying that the D4 manages to achieve the same high ISO performance as the D3s, but with a higher pixel count. Thus, the 1D X would need to be between 3/4 and 1 stop better than the 5D MkII to match the D4.

3. I've got a suspicion that the D3s 'cooks' its raw files at high ISO, based upon the SNR 18% graph from DXO Mark having a strange kink in it after ~ISO 9787 (also note the performance of the D3).

Anyone want to check these calculations?

checking the other graphs, the same pattern holds up in the other graphs -- the D3s widens the gap at high ISOs in measures such as dynamic range, tonal range and color sensitivity.

 

[briansquibb]

The Nikon D3s being only 3/4 of a stop better then the 5DmkII seems low. If your calculation are right, the hill the climb is much less then I thought...but I recall seeing some very high ISO sample from the D3s that just destroy the 5DmkII at 12800 and above. Not sure how these charts works, but something is weird here...

The 5DII seems to manage 3200 well, anything more and IQ/noise gets worse quickly.

Not so with the 1D4 which seems to top out at 12800

 

[Mt Spokane Photography]

A few points:

1. According to DXO Mark, the D3s reaches its "acceptable" signal to noise value of 30dB at 3253 ISO (whilst keeping dynamic range at >9 EV and colour depth at 18bits), whereas the 5D MkII achieves this at 1815 ISO. So 3253/1815 = ~1.79; i.e. the D3s is 79% better than the 5D MkII, which is basically just over 3/4 of a stop.

2. Nikon is implying that the D4 manages to achieve the same high ISO performance as the D3s, but with a higher pixel count. Thus, the 1D X would need to be between 3/4 and 1 stop better than the 5D MkII to match the D4.

3. I've got a suspicion that the D3s 'cooks' its raw files at high ISO, based upon the SNR 18% graph from DXO Mark having a strange kink in it after ~ISO 9787 (also note the performance of the D3).

Anyone want to check these calculations?

checking the other graphs, the same pattern holds up in the other graphs -- the D3s widens the gap at high ISOs in measures such as dynamic range, tonal range and color sensitivity.

There is little doubt, its a trade off, high resolution and ability to crop versus low resolution and better high ISO performance. I wonder how a 22mp 5D MK III at $2800 will compare against the 36MP D800 at 3000.

There will be those who like the higher resolution, and those who like the better high ISO sensitivity, its either-or.

 

[JR]

The Nikon D3s being only 3/4 of a stop better then the 5DmkII seems low. If your calculation are right, the hill the climb is much less then I thought...but I recall seeing some very high ISO sample from the D3s that just destroy the 5DmkII at 12800 and above. Not sure how these charts works, but something is weird here...

one thing to keep in mind is that this test only shows how well the two cameras do at 3200 and 1800 ISO. DxOMark tests how high you can crank the ISO and maintain a certain level of IQ (I think defined in terms of dynamic range and noise)

It doesn't say anything about how well they hold up above that (how much they lose if you turn ISO higher), and there's no reason the curves should be parallel. One example is that medium format cameras do very well at low ISO but not as well at high ISO as full frames.

I never realized that. I guess it goes to show how we need to learn to read behind spec sometime to get the full story! Indeed Nikon must do something to the very High ISO to make it look good - I just hope someone at Canon figured this out as well for their next line of EOS bodies !!!

 

[JR]

The Nikon D3s being only 3/4 of a stop better then the 5DmkII seems low. If your calculation are right, the hill the climb is much less then I thought...but I recall seeing some very high ISO sample from the D3s that just destroy the 5DmkII at 12800 and above. Not sure how these charts works, but something is weird here...

The 5DII seems to manage 3200 well, anything more and IQ/noise gets worse quickly.

Not so with the 1D4 which seems to top out at 12800

It is so interesting you mention this Brian because while indeed everyone claim the 5DmkII to be better at low light then the 1DIV (I guess based on DxO chart for example) I always found the 1DIV (whenever I can borough one) to be very good at very high ISO. Because I dont own both bodies, I never knew if it was the shooting environment at that time that made it look good (I am refering to ISO 6400 and above) or because it was the 1DIV...

 

[JR]

There is little doubt, its a trade off, high resolution and ability to crop versus low resolution and better high ISO performance. I wonder how a 22mp 5D MK III at $2800 will compare against the 36MP D800 at 3000.

There will be those who like the higher resolution, and those who like the better high ISO sensitivity, its either-or.

At one point Nikon Rumors had some ISO sample of the D800 on their site. While it was unclear if they were real of fake, the sample looked much worst then the current 5DmkII at ISO performance. For my use 21MP is more then enought for my croping need, so my trade off will be better ISO performance once I have a machine with 18-21MP...

 

[Woody]

Personally, I just process comparable RAW images from the cameras of interest using the best software out there and compare afterwards. That is most relevant for me as it is based on software I like to work with.

 

[Ricku]

Improvement only in jpeg files does not matter to me and MANY other photographers, because we only shoot in RAW.

 

[aj1575]

canon will get f£$ked by interviews like this one, well all is dommed now, good luck guys im jumping to the Nikon boat.

The grass is always greener on the other side....

....or do you think that a chance from brand A to brand B will really improve your pictures?

 

[briansquibb]

The Nikon D3s being only 3/4 of a stop better then the 5DmkII seems low. If your calculation are right, the hill the climb is much less then I thought...but I recall seeing some very high ISO sample from the D3s that just destroy the 5DmkII at 12800 and above. Not sure how these charts works, but something is weird here...

The 5DII seems to manage 3200 well, anything more and IQ/noise gets worse quickly.

Not so with the 1D4 which seems to top out at 12800

It is so interesting you mention this Brian because while indeed everyone claim the 5DmkII to be better at low light then the 1DIV (I guess based on DxO chart for example) I always found the 1DIV (whenever I can borough one) to be very good at very high ISO. Because I dont own both bodies, I never knew if it was the shooting environment at that time that made it look good (I am refering to ISO 6400 and above) or because it was the 1DIV...

I think the 5DII focusses better in very low light but the image is bettter from the 1D4 at 6400 and 12800 (which is not that low, just a room light level)