High ISO, 5DIII vs 5DSR

[aftab]

How does 5DSR compare with world's most popular DSLR 5DIII in high ISO? Is it just a one trick pony?

First thing I noticed when comparing 5DSR with 5DIII is that 5DIII images appear out of focus when viewed at 100%, this becomes more striking when 5DSR image is downsized to 5DIII size. By sharpening 5DIII RAW image the difference could be reduced, but this becomes a problem in high ISO.

All images in this test were shot RAW and processed in ACR/CS6. Couple of things before I show the comparison samples.

Using same exposure settings 5DIII images appear noticeably brighter irrespective of ISO used. I adjusted the brightness slider in ACR to bring them as close as possible.

At ACR's default chroma NR (25) some colors in 5DSR fades (looks washed out) while 5DIII does fine. Setting chroma NR to 10 for 5DSR retains the colors while removing all color noise. For 5DIII though, Chroma NR of 10 is not often enough to remove all color noise.

ISO3200

5DIII, 100% crop, no NR


5DIII, 100% crop, no NR

5DSR, 100% crop after downsizing to 5DIII size, no NR


5DSR, 100% crop after downsizing to 5DIII size, no NR

Now, if you look at the larger sizes by clicking on the pictures you will notice that 5DSR actually shows more noise even after downsizing. But 5DSR also shows significantly more detail after downsizing (look at the top half of the bill). So, even after NR 5DSR wins clearly.


5DSR, NR

I didn't try NR on 5DIII for this sample, it was too soft to apply NR. This is how 5DSR looked at 100% before size reduction and NR.



Another example at ISO3200. Previous shots were taken with AF. This time I decided to go manual AF in live view to eliminate any AF error.

5DIII, 100%, no NR


5DIII, 100%, no NR

5DSR 100% after downsizing to 5DIII size, no NR



Now, both files processed.


5DIII


5DSR

Look at larger sizes by clicking on the pictures. Same story here, but in this example in the absence of very fine detail 5DIII does better compared to previous example.

ISO6400

Again manual focus in live view with magnification, 100% crops, 5DSR downsized.

5DIII, no NR


5DIII, no NR

5DSR, no NR


5DIII, no NR

5DIII, NR


5DIII, NR

5DSR, NR


5DSR, NR

So, up to ISO 6400 5DSR does better than 5DIII if you process the files. At higher ISOs it is almost certain that you would do some processing.

I can post the RAW files if anyone interested.

--

Catch45

Explore Catch45’s 200 photos on Flickr!

www.flickr.com

 

[J A C S]

This is misfocused

5DIII, 100% crop, no NR

On the right: is that real texture or aliasing?
The globe: misfocused again.

5DIII, no NR

 

[ASR45]

This is misfocused

I agree, in a nutshell the 5D3 is better at ISO simple factor less MPs.

5DIII, 100% crop, no NR

On the right: is that real texture or aliasing?
The globe: misfocused again.

5DIII, no NR

--
Alan.
Great photography is about depth of feeling, not depth of field.
- Peter Adams
You don't take a photograph, you make it. - Ansel Adams.

 

[aftab]

This is misfocused

That's what it looks like.

These are AF, 100% crop, RAW, no sharpening.

5DIII



5DSR



5DIII



5DSR

On the right: is that real texture or aliasing?

I am not able to see the pattern on the box. Must be aliasing.

The globe: misfocused again.

See above. In more than 50 side by side comparison with 3 different lenses, it is the same story.

Catch45

Explore Catch45’s 200 photos on Flickr!

www.flickr.com

 

[aftab]

This is misfocused

I agree, in a nutshell the 5D3 is better at ISO simple factor less MPs.

Not according to my test and not according to Bryan's test.

http://www.the-digital-picture.com/Reviews/Canon-EOS-5Ds.aspx

This is what he had to say: "When compared at native resolutions, 5Ds images are noisier than 5D III images. The differences, especially at higher ISO settings, are less than 1 stop. Down-sized to 5D III pixel dimension (using DPP, see "Standard Down-Sized to 5D III" in noise tool), 5Ds noise levels are essentially equal to full frame 5D III noise levels and even slightly better at the highest ISO settings. So, while Canon is not promoting this camera for its low light capabilities, I see it as one of the best options available with output size being comparable."

(emphasis mine)

 

[J A C S]

See above. In more than 50 side by side comparison with 3 different lenses, it is the same story.

Loading…

www.dpreview.com

Loading…

www.dpreview.com

 

[aftab]

See above. In more than 50 side by side comparison with 3 different lenses, it is the same story.

http://www.dpreview.com/reviews/ima...1&x=-0.6533486401550752&y=0.25170205505085574

http://www.dpreview.com/reviews/ima...=1&x=-0.5409828954710997&y=0.3138503671421684

In both cases 5DSR images appear sharper. Here is one more (I deleted most test shots).

You know which camera took which picture (100% crops). One will need some sharpening and the other won't.





In low ISO situations 5DIII images can be easily sharpened, in high ISOs that's not possible in most cases



--

Catch45

Explore Catch45’s 200 photos on Flickr!

www.flickr.com

 

[aftab]

Two full scenes from 5DIII, ISO100 and ISO3200. Detail it captures in both situations is quite remarkable (we were looking at a very small area). Also, look how clean the ISO3200 image is when looked at 3000 pixels wide (no NR). So, I am not putting down 5DIII. It is just that AA filter cancellation makes images look sharper and captures finer detail at all ISOs. Higher MP may also play a role even after size is normalized. They work in favor of 5DSR in all ISOs. I don't know if S version will have the same advantage.















--

Catch45

Explore Catch45’s 200 photos on Flickr!

www.flickr.com

 

[marshwader]

Your tests look pretty conclusive to me. But: Just as I had finally come down on the side of getting the "R" the aliasing raises its ugly head on the shapes packet. How much of a problem is it really. Shame that there are so few 5ds images about to really judge how much sharpness is lost.

Anyway. Thanks very much for doing and publishing all these tests.

 

[Victor Engel]

Your tests look pretty conclusive to me. But: Just as I had finally come down on the side of getting the "R" the aliasing raises its ugly head on the shapes packet. How much of a problem is it really. Shame that there are so few 5ds images about to really judge how much sharpness is lost.

Have you not seen the comparison shots here at dpreview? To my eye, there's not much difference in sharpness between the two, especially if you apply appropriate sharpening to the 5DS shots. In fact, the AA filter appears to be a weak filter because moire does show up on some 5DS shots.

Anyway. Thanks very much for doing and publishing all these tests.

 

[aftab]

Your tests look pretty conclusive to me. But: Just as I had finally come down on the side of getting the "R" the aliasing raises its ugly head on the shapes packet. How much of a problem is it really. Shame that there are so few 5ds images about to really judge how much sharpness is lost.

Judging by DPR's comparison, sharpness difference between S and R version is extremely low. Same goes for aliasing too, they show up in both versions, in R version they just appear sharper. The pattern on the Shapes packet is interesting, reminds me of the pattern we see in IR samples of some AA less camera. This is from Pentax 645D

Anyway. Thanks very much for doing and publishing all these tests.

Thanks.



--

Catch45

Explore Catch45’s 200 photos on Flickr!

www.flickr.com

 

[RedFox88]

Just how are you reducing the 5Dsr pics? Bicubic sharper? Sharpening after resize?

any sharpening to the 5D3 pics?

In my opinion because of resizing the larger pics, pics from both cameras need to be sharpened to show each at its best.

 

[aftab]

Just how are you reducing the 5Dsr pics? Bicubic sharper? Sharpening after resize?

Bicubic automatic in CS6.

any sharpening to the 5D3 pics?

'Shapes' ISO3200 samples and ISO6400 sample are sharpened after NR. Outdoor pics from both cameras are not sharpened.

In my opinion because of resizing the larger pics, pics from both cameras need to be sharpened to show each at its best.

Yes. After proper sharpening/processing downsized images become indistinguishable unless viewed at 100%.

 

[marshwader]

Thank you for your replies. Over the last few weeks I've stared at every 5ds and dsR sample I could lay my hands on to understand the subtle differences in sharpness. aliasing etc. and I've changed my mind a few times as to which to go for. Side by side comparisons of the same (landscape) scene would be great but pretty unrealistic at this point, I guess. I'll plumb for either one within the next week as they have only just become available here in the UK where I currently work and also elsewhere in Europe. I'll probably get it from Germany where it is very much cheaper. God Knows why. The DR question doesn't trouble me so much as most of my jobs are in controlled lighting set-ups.The landscapes I also like portraying don't involve plastic tulips revealed in deeply shadowy fields and you can only do so many sun stars before you feel heartily sick of them.

Thanks again for your very involved work. I am also very impressed by Horshacks mega-lens test.

You are all doing fantastic work for the community here.

 

[aftab]

Thank you for your replies. Over the last few weeks I've stared at every 5ds and dsR sample I could lay my hands on to understand the subtle differences in sharpness. aliasing etc. and I've changed my mind a few times as to which to go for. Side by side comparisons of the same (landscape) scene would be great but pretty unrealistic at this point, I guess. I'll plumb for either one within the next week as they have only just become available here in the UK where I currently work and also elsewhere in Europe. I'll probably get it from Germany where it is very much cheaper. God Knows why. The DR question doesn't trouble me so much as most of my jobs are in controlled lighting set-ups.The landscapes I also like portraying don't involve plastic tulips revealed in deeply shadowy fields and you can only do so many sun stars before you feel heartily sick of them.

Thanks again for your very involved work. I am also very impressed by Horshacks mega-lens test.

You are all doing fantastic work for the community here.

Thanks for the kind words.

Yeah, Horshack's mega-lens test is really mega and highly admirable.

I always worry about my tests though. If they are not accurate enough they might mislead few people in drawing wrong conclusions. So, I might repeat some of my tests.

Many people had your dilemma with D800 and D800E. At the end Nikon sold a lot more D800. But interestingly enough, they decided to go AA less.

 

[RedFox88]

Just how are you reducing the 5Dsr pics? Bicubic sharper? Sharpening after resize?

Bicubic automatic in CS6.

any sharpening to the 5D3 pics?

'Shapes' ISO3200 samples and ISO6400 sample are sharpened after NR. Outdoor pics from both cameras are not sharpened.

In my opinion because of resizing the larger pics, pics from both cameras need to be sharpened to show each at its best.

Yes. After proper sharpening/processing downsized images become indistinguishable unless viewed at 100%.

 

[BlueRay2]

Aftab, thanks for your work! the 5Dsr is completely a winner and impressive, makes 5DIII look like an old foot-mat ;-)

 

[aftab]

As others have mentioned some of your 5D3 pics are out of focus.

This is from IR. 5DSR resized to 5DIII size. No sharpening, no NR, only ACR's default chroma NR.

5DIII





5DSR



Look at the large size and see the difference. 5DIII appear out of focus.



--

Catch45

Explore Catch45’s 200 photos on Flickr!

www.flickr.com

 

[aftab]

Aftab, thanks for your work! the 5Dsr is completely a winner and impressive, makes 5DIII look like an old foot-mat ;-)

Thanks. I won't call it a foot-mat though.

 

[J A C S]

I always worry about my tests though. If they are not accurate enough they might mislead few people in drawing wrong conclusions. So, I might repeat some of my tests.

Why don't you use both in actual low light, artificial lighting, handheld, no LV, with close to critical SS? Also, no closeups.