HUGE difference between the 5d Mark 3 and D800 RAWs... I mean... HUUGEE

[tonyp]

http://forums.dpreview.com/forums/read.asp?forum=1032&message=40912195



Can we stop now. It's pretty much a wash....

 

[SpartanWarrior]

Huge?? lol hardly download the RAW files and compare, you will see the 5D III beats the D800 at ISO and the detail on the D800 is not that much better.

 

[tonyp]

I guess my sarcasm was not noticed... 8)

 

[SpartanWarrior]

Sorry, they both will be great though

 

[jalbfb]

I'm one of those who pre-ordered the 5D3 and am anxiously awaiting it's arrival. I have invested a lot in good Canon glass and for me there was never a thought of switching to Nikon. I was only hoping the 5d3 would be an improvement over the 5DII which I bought last year when the tragedy hit Japan and I figured it would take a year for them to recover. Also, the higher MP count was a non-issue for me. i actually favored the lower count. It looks like BOTH cameras are going to be great additions for those making a camera selection. They give everyone choices and options to stay with either Nikon or Canon, or switch systems. In the end, it's a choice you have to make based on your individual camera needs/uses, how much you have already invested, can you afford to switch or can you even afford the camera. Competition is great and fuels better equipment for all of us to choose from. I also provide fodder for the various blogs which i find both entertaining and educational to follow. I've learned a lot from just following the various threads about camera/photo technology and capabilities and some of the highly skilled and educated camera buffs and pros. IMO, everyone wins.

 

[Maui5150]

Correct me if I am wrong, but what exactly does: "5D3 was upsided to 36 MP via Photoshop Bicubic" mean???

My interpretation is they took the Canon 22MP file, and in PhotoShop did an Image Resize using bicubic to be equal in size to 36 MP

I have NEVER been able to resize a photo up by 20% or more and not introduce defects, loss of sharpness, etc.

 

[awinphoto]

Correct me if I am wrong, but what exactly does: "5D3 was upsided to 36 MP via Photoshop Bicubic" mean???

My interpretation is they took the Canon 22MP file, and in PhotoShop did an Image Resize using bicubic to be equal in size to 36 MP

I have NEVER been able to resize a photo up by 20% or more and not introduce defects, loss of sharpness, etc.

That appears to be what happened... Puts the 5d3 at a slight disadvantage but for arguements sake, it shows how well it stands up when compared toe to toe.

 

[psolberg]

Huge?? lol hardly download the RAW files and compare, you will see the 5D III beats the D800 at ISO and the detail on the D800 is not that much better.

http://forums.dpreview.com/forums/read.asp?forum=1032&message=40912195



Can we stop now. It's pretty much a wash....

it was posted already but IMO, the much higher detail of the D800 is really noticeable.
http://www.imaging-resource.com/news/2012/03/14/nikon-d800-test-shots-captured-and-posted

Huge?? lol hardly download the RAW files and compare, you will see the 5D III beats the D800 at ISO and the detail on the D800 is not that much better.

I'd dissagree. I say it is noticeably better. Off course the 5DIII can go higher on the ISO with less noise. Downsampling the D800 raw to 22MP does but it in 5DIII terrytory though.

overall I think the D800 remains the better camera by far to its purpose: studio/landscape. I'm not sure high ISO tests really get what the camera is all about.

 

[Joe J]

Huge?? lol hardly download the RAW files and compare, you will see the 5D III beats the D800 at ISO and the detail on the D800 is not that much better.

I'd dissagree. I say it is noticeably better. Off course the 5DIII can go higher on the ISO with less noise. Downsampling the D800 raw to 22MP does but it in 5DIII terrytory though.
overall I think the D800 remains the better camera by far to its purpose: studio/landscape. I'm not sure high ISO tests really get what the camera is all about.

+1.
If you actually look at the photos without Canon-colored glasses, there is a significant difference in detail and clarity, particularly with the 1st, 3rd and 4th ones. Implying that lack of detail in one image compared to another is not much different just makes one's analysis of them have zero credibility.
I could care less about either camera (1DX is really the only new camera model for professionals to consider anyway), and I'm a rejuvenated Canon supporter thanks to the 1DX announcement, but that doesn't mean one should overlook the facts.

 

[HurtinMinorKey]

What lenses were used? I would imagine the differences in lenses would be as likely a suspect for perceived differences as anything else.

 

[awinphoto]

Huge?? lol hardly download the RAW files and compare, you will see the 5D III beats the D800 at ISO and the detail on the D800 is not that much better.

I'd dissagree. I say it is noticeably better. Off course the 5DIII can go higher on the ISO with less noise. Downsampling the D800 raw to 22MP does but it in 5DIII terrytory though.
overall I think the D800 remains the better camera by far to its purpose: studio/landscape. I'm not sure high ISO tests really get what the camera is all about.

+1.
If you actually look at the photos without Canon-colored glasses, there is a significant difference in detail and clarity, particularly with the 1st, 3rd and 4th ones. Implying that lack of detail in one image compared to another is not much different just makes one's analysis of them have zero credibility.
I could care less about either camera (1DX is really the only new camera model for professionals to consider anyway), and I'm a rejuvenated Canon supporter thanks to the 1DX announcement, but that doesn't mean one should overlook the facts.

There was plenty of detail in the nikons, maybe even edging canon, but also you have to take into account they enlarged the 5d3 files to match those of the D800, and it is unknown, however doubtful, if they sharpened the file post enlargement or kept the file as is... Assuming they didn't, they basically took the 5d file, englarged it and put the file through a disadvantage, yet, it also shows less noise and details wise, held it's own vs the D800... now lets say you take the D800, reduce it to meet the 5d3, the D800 probably is a lot closer in noise, but details would probably be a lot closer if not the same. This is a lot like when the 5d2 came out and review websites would either enlarge the D700's files to match the 5d2, which came out horrible compared, or took the 5d2 and reduced it to the D700, which looked pretty damn close. Of course there will be hurt feels by one side or the other, but this is the worst case situation for the 5d3 and it holds it's own. Of course, 90% of 5d3 shooters will likely use the files as is and few will take those files, enlarge the files to that of the D800 or larger, unless doing such for print, but then likely they will add print sharpening which will make the files look pretty good. Take the test for what it is and the test looks pretty good no matter what side of the fanclub you belong to.

 

[Maui5150]

Huge?? lol hardly download the RAW files and compare, you will see the 5D III beats the D800 at ISO and the detail on the D800 is not that much better.

I'd dissagree. I say it is noticeably better. Off course the 5DIII can go higher on the ISO with less noise. Downsampling the D800 raw to 22MP does but it in 5DIII terrytory though.
overall I think the D800 remains the better camera by far to its purpose: studio/landscape. I'm not sure high ISO tests really get what the camera is all about.

+1.
If you actually look at the photos without Canon-colored glasses, there is a significant difference in detail and clarity, particularly with the 1st, 3rd and 4th ones. Implying that lack of detail in one image compared to another is not much different just makes one's analysis of them have zero credibility.
I could care less about either camera (1DX is really the only new camera model for professionals to consider anyway), and I'm a rejuvenated Canon supporter thanks to the 1DX announcement, but that doesn't mean one should overlook the facts.

Really?

The Canon images were sampled up.

Quick test.

Take an image from a Nikon D4 and Bicubic resample it to 36MP and then compare it to the D800.

Which is sharper at ISO 100?

By this logic the D800 is a vastly sharper/better camera than the D4

You cannot take a 22MP image, resample / increase the image size to 36 MP AND NOT have a loss in quality.

In fact, you could probably take a D4 16MP image at ISO 100 and Upscale it to 21MP and compare to a 5D MKII and see how the 5DMKII is a SHARPER IMAGE than the D4?

You might as well compare the RAW from one to a JPG from the other.

Why not shoot the Nikon in Medium instead of Large for the file size which would be more indicative of a comparable resolution???

Ahhh... because maybe that would not show the results the bias was tending to show.

Raw is Raw... You can not take images from 1 that is pretty much from Raw and compared to another that is upsampled by close to 30% and expect to make a true judge of sharpness.

 

[unfocused]

I have to say that I was convinced until someone pointed out the upsampling of the image. I wonder why that is necessary.

It seems that if you wanted to do a fair comparison between two cameras you would take identical sample images, and then examine the images at various magnifications to see at what point, if any, the differences become noticeable.

Okay, I suppose the person making these tests was trying to show the difference between using the original file and creating an upsampled version at the same resolution. But, doesn't that become as much a test of upsampling software as it does the cameras?

Frankly, all this amateur pixel peeping, followed by flame wars on thread after thread, is getting a bit boring.

The more I read, the more convinced I become that the differences between comparable models of Canon and Nikon are minor and mainly involved very narrow functional and design nuances.

Unless you are in that tiny, tiny percentage of photographers for whom these differences actually matter it seems like there are two rational ways to decide. You can count up all your equipment and calculate the cost of switching or you can stick pictures of both cameras on the wall, blindfold yourself and then throw a dart at the wall, picking the camera that the dart comes closest to.

Either way, it's unlikely to affect your pictures in any meaningful or measurable fashion.

 

[stve]

Yeah it's a massive difference in favour of the D800

http://masters.galleries.dpreview.com.s3.amazonaws.com/1813619.jpg?AWSAccessKeyId=14Y3MT0G2J4Y72K3ZXR2&Expires=1331830990&Signature=ny7zRykvg%2fh8vDdLDfv%2fGV5A4k0%3d

 

[Loswr]

You can not take images from 1 that is pretty much from Raw and compared to another that is upsampled by close to 30% and expect to make a true judge of sharpness.

Sure you can. All you have to do is decide ahead of time what 'conclusion' you want your 'test' to show, and then design your test accordingly. :

 

[smirkypants]

I wonder what would happen if you had to crop away 50% of the photo and then upsample both. I have been known to crop a wide photo tall and vice versa. You lose at least half the photo. Say then after that you wanted to do a 20x30 print (which I do quite frequently), where would we stand? Do you suppose the minor differences would still be minor?

This is an honest question, not a "sez you!"

 

[tonyp]

Shoot the scene both ratios... problem solved and keep the true MPs

 

[Maui5150]

You can not take images from 1 that is pretty much from Raw and compared to another that is upsampled by close to 30% and expect to make a true judge of sharpness.

Sure you can. All you have to do is decide ahead of time what 'conclusion' you want your 'test' to show, and then design your test accordingly. :

Exactly.

I can understand some of the initial premise of the upsizing, i.e. showing the same size crop so the size of an object shown in the crop is the same size, but this just does not work.

If anything, what it tells me is either:

1) The author is intentionally being deceptive

- or -

2) They really do not understand digital files, resolution and comparison.

To me the proper way of doing this comparison has to be raw to raw with an understanding that the the Nikon image will obviously be blown up a bit, but looking at a 500 x 500 pixel representation of the Canon Raw and then comparing to the Nikon Raw, one can look at the edges, CA, noise, etc.

By the same token, another poster raised an excellent point that the lenses used will also make a big difference as I expect the quality of say a 70-200 L2.8 IS II will produce a different image than a 24-105 F/4.

By that degree, Camera to Camera comparisons can be very selective between different brands, where as at least within a brand, lens performance and difference can be mitigated.

 

[awinphoto]

I wonder what would happen if you had to crop away 50% of the photo and then upsample both. I have been known to crop a wide photo tall and vice versa. You lose at least half the photo. Say then after that you wanted to do a 20x30 print (which I do quite frequently), where would we stand? Do you suppose the minor differences would still be minor?

This is an honest question, not a "sez you!"

Theoretically, if both are cropped exactly at 50% each, the D800 would still have the bigger file and then when enlarged, the D800 would have the edge, but if you cropped the D800 in similar size to the 5d3 so they are both x pixels by y pixels, and enlarge both to similar sizes, then maybe canon has the edge because it would have less manipulations/artifacts, but would be pretty close

 

[Loswr]

To me the proper way of doing this comparison has to be raw to raw with an understanding that the the Nikon image will obviously be blown up a bit, but looking at a 500 x 500 pixel representation of the Canon Raw and then comparing to the Nikon Raw, one can look at the edges, CA, noise, etc.

Alternatively, upsample both to 40 MP using something like Genuine Fractals, where the degree of upscaling has less of an impact than the presence/absence of the manipulation.