DxOmark Canon 5DIII

[marcmccalmont]

Finally DxO canon 5DIII review! 81

[shadowblade]

Pretty much expected, really. The 5D3 was never going to compete on RAW image quality.

Canon really needs to lift its game in the sensor department, or risk losing the entire landscape and studio crowd. The current offerings are nice for journalists and event photographers, but not so nice if image quality is your paramount consideration.

A real pity, given their lens lineup, and the fact that, just a few years ago, the 1Ds3 (and subsequent 5D2) were top of the league.

[BernardLanguillier]

There must be some high level manager in Canon Imaging who decided some years ago that DxOMark was not a relevant factor in their extended eco-system. Why should they have cared about a small French Start up with no track record?

The business cost of this decision may be extremely high.

Cheers,
Bernard

[Bart_van_der_Wolf]

Pretty much expected, really. The 5D3 was never going to compete on RAW image quality.

Indeed, and I kind of predicted the Dynamic Range performance (except for some missing data to put it into numbers) here. Although I must say that the DxO figures for (screen) DR are about 1/3rd of a stop worse than what I expected (when assuming a clipping level of around 16000 in 14-bit data numbers). The almost level performance from ISO 100 to ISO 800 (at unity gain) was to be expected, as was the deviating-from-the-trend response for the 3 highest ISO settings.

Cheers,
Bart

[marcmccalmont]

There must be some high level manager in Canon Imaging who decided some years ago that DxOMark was not a relevant factor in their extended eco-system. Why should they have cared about a small French Start up with no track record?

The business cost of this decision may be extremely high.

Cheers,
Bernard

Bernard
The manager thought the R&D cost to develop a new sensor was way too high so he had the marketing  department talk a dealer into posting YouTube videos stating his camera is "as good as the competition" even better in low light conditions, then so it didn't look too biased lets compare the competitor to a Hassy and state the competitor "only has 16 skin tones not 27" cost a lot less than R&D, smart business man "eh"
Marc

ps funny how DxO compared the D800 to every camera out there yet in the 5DIII review they only compared it to it's predecessor? As a Canon lover I'm really disappointed. 

[deejjjaaaa]

incidentally a "news" article = http://news.cnet.com/8301-11386_3-57415777-76/how-dxo-labs-tests-hot-cameras-like-canons-latest-slr/

[Ray]

Marc, you must have been checking for those results daily to notice them so quickly. 

Just in case anyone is confused about the noise differences betwee the 5D2 and 5D3, as shown on the DXOMark graphs, I'll try and summarize them.

(1)ISO sensitivities are about the same for both cameras at lower ISOs, the 5D3 being very slightly more sensitive to an insignificant degree. However, such differences seem to increase as ISO increases, so by ISO 25,600 the 5D3 is about 1/4 of a stop more sensitive.

What this means in practice is that when using the same shutter speed at ISO 25,600 with both cameras, the 5D2 sensor will receive 1/4 of a stop less exposure (in terms of its full-well capacity) and will therefore have 1/4 of a stop more noise, and 0.25EV less DR, irrespective of any other differences

However, there are other differences. At the manufacturer's nominated ISO of 25,600, the 5D3 has very close to 1 stop lower noise at 18% grey (the midtones), and very close to 1 stop better DR. Add the 1/4th of a stop difference in ISO sentivities and we get slightly more than 1 stop better DR and slightly more than 1 stop lower midtone noise for the 5D3, at the same shutter speed, with the same lens used at the same aperture.

However, if different lenses are used, then all bets are off. Using different lenses with both cameras, even at the same F/stop and same focal length, using the two cameras at ISO 25,600 and at the same shutter speed, the sensors could either have close to the same performance, or the gap could be widened to as much as a 2 stop difference, depending on the T-stop values of the two lenses.

(2) At the more commonly used ISOs, from base up to ISO 1600, there's an insignificant difference between the two cameras with respect to SNR, DR, Tonal Range and Color Sensitivity. At extreme pixel-peeping levels, one might be able to discern very slightly less midtone noise in the 5D3 image, but I doubt it.

Up to ISO 1600, DR for both cameras is virtually identical, with a very slight edge going to the 5D2, which is surprising. However, such difference are probably within the fluctuation range due to QC differences between different units of the same camera model, and any imprecision in the testing procedures. Nothing's perfect.

(3) Bottom line: any improvement in the 5D3 sensor performance (regarding noise and DR compared with the 5D2), seems to be confined to very high ISOs above ISO 1600. The 12mp D3s still has the edge at high ISO, having on average about 1/2 a stop better DR at the same print size, at high ISO.

Hope this is all crystal clear. 

[julius0377]

Its weird that the 5DmkIII gets only 2293 score in low light iso tests vx. the D800's 2853, if one looks at the individual pr. pixel measurements the 5DmkIII has the edge in both DR, signal to noise as well as tonal range, except for a slight disadvantage in color sensitivity. Odd that these measurements should lead to such a score difference for the low light iso?

Interesting to see that the D800 outclasses the 5DmkIII in dynamic range up to about iso800, but after iso1200 the 5DmkIII gains a slight and increasing edge on the D800 up to maximum iso. (Using the DXOmark "screen" setting for measurements.)

Maybe DXOmark puts a lot more emphasis on the print section of the measurements, but would this not be slightly misleading, as the application would only be relevant at certain large print sizes?

[DaveCurtis]

The measurements reported are as expected, rather disappointing. They are also in aggreement with my 'real world' observations.

I have spent several days shooting my 1DS3 and my new 5D3 side by side. They are essentially even stevens to about ISO 1600. At 3200 there is a small gain to the 5D3. However my problem is I would shoot the majority of my image between ISO 100 - 1600 and I sure most photographers would be the same.

What has Canon been doing for they last 3.5 years ? Perhaps spending all their resourses on their new FF video cameras ?

Credit where credit is due though, the AF works brilliantly.

[shadowblade]

Maybe Canon knows about some upcoming, sub-$10k medium-format bodies with good UWA lenses to go with them, thus rendering DSLRs irrelevant to the studio and landscape crowds - hence, they're concentrating on event/wedding/sports photographers.

[BernardLanguillier]

Maybe Canon knows about some upcoming, sub-$10k medium-format bodies with good UWA lenses to go with them, thus rendering DSLRs irrelevant to the studio and landscape crowds - hence, they're concentrating on event/wedding/sports photographers.

Partial understanding of market expectations, lukewarm investment in core technologies and focus on video seem more likely explanations to me.

Cheers,
Bernard

[hjulenissen]

Partial understanding of market expectations, lukewarm investment in core technologies and focus on video seem more likely explanations to me.

1. Most photography review people, and most customers thinks that smaller pixels means worse image quality. Many of those will buy/recommend the 5Dmk3 over the D800 based on this theory alone, disregarding the actual image quality.
2. Many photographers and reviewers desire "better high ISO image quality". They may not know or care about DR at ISO100.
3. Most buyers of the 5Dmk3 (or pretty much any model) will be wealthy hobbyists, not technically inclined photographes, and certainly not image scientists.
4. Video, in-camera JPEG/HDR, etc is there for a reason: only some percentage of customers have the inclination to exploit the sensor to its fullest for still-images, and those may be the only ones who prioritize sensor performance over everything else.

Being invested in the Canon "ecosystem", I think this is kind of sad. Hopefully, Canon will return with a killer sensor within the 2-3 years that I expect to upgrade.

-h

[MrSmith]

why don't they test the 5dIII at 160/320/640 iso where it has more dynamic range and less noise.
i do find this whole dxo thing a bit odd, a measurebators resource for the weak minded. do people really wait for the score or what some bloke on a blog says about a camera? actually don't answer that 

[BJL]

Actually, the 5D3 seems to have just one disadvantage beyond the obvious one of lower resolution: as exposire index is increased (lower ISO, to those who insist on confusing exposure index with sensitivity by using that highly ambiguous name for both of them), Canon loses some of the shadoe performance that the sensor is clearly capable of, apparenly because with wells close tp full, it has to use an amplification so low that noose floor of subsequent signal path is several stops higher than the noise floor of the signal from the sensor itself. As soon as the light level is low enough (higher shutter speed), the performance at equalized resolution("print") looks very similar. Actually, this makes me wonder if in the measurements at higher EI, we are seeing close to the optimal behavior set by photon noise.

Anyway, Canon might have judged that only a small proportion of potential customers have a real need for more than about a 2000:1 SNR (11 stops of engineering DR), since anything beyond that is going to be visible only with substantial manupulations of tonal compression (exteme low contrast printing and/or heavy dodging and burning, in the old language). And for that extreme, Canon has perhaps judged that, for now, in-camera HDR is an adequate partial solution, and the extra cost of developing and deploying a substantially new sensor technology like column-parallel ADC is not cost-effective.


By the way, column parallel ADC is not a Sony exclusive: it has been used by other companies such as Samsung in video sensors. I wonder if the new Samsung 20MP "APC-C" format sensor in the new NX bodies uses column-parallel ADC? Which by tue way makes me think it is only a matter of when Canon deploys a next generation sensor technology like "CPADC", not any possibility that Canon is out of the game.

[ejmartin]

why don't they test the 5dIII at 160/320/640 iso where it has more dynamic range and less noise.

Because it (a) doesn't have more DR at these ISO's, and (b) doesn't have less noise at these ISO's.  But then, you might be fooled into thinking so if you didn't know something about how Canons work, by (surprise!) measuring their properties.

These ISO's are achieved by digital compression of the signal from a different analog gain; the noise is less in digital numbers but no better when denominated in photon equivalents.  The DR range is the same as it is for the analog ISO gain that these oddball ISO's are derived from by digital scaling.

[LKaven]

why don't they test the 5dIII at 160/320/640 iso where it has more dynamic range and less noise.
i do find this whole dxo thing a bit odd, a measurebators resource for the weak minded. do people really wait for the score or what some bloke on a blog says about a camera?

While tests don't always measure such things as pattern noise and thermal noise, they do give a very good idea of performance in key areas.  Since I'm always looking to push low light photography, measurements have been a useful guide.  More than that even, the knowledge of people who do study these sensors, some of whom are writing in this thread, has been enormously useful.  Certainly the rest comes in practice.

[EgillBjarki]

I respect the tests of DxOMark, it seems clear that Nikon made a really solid update, with Canon only making mild upgrades from Mark II.

How ever, this does not mean that Nikon users will be the only photographers producing good landscape work in the coming years.

I bought the 5D Mark III yesterday, now my trusted Mark II is my backup body. Before I used D700, in 2010 I moved to 5D Mark II for resolution and video. Right now the argument could be made for me to go to Nikon for the same reason.

In the end I don't think any of this matters much... DR is never a problem with portraits when I use strobes. When I need DR, I can work around it and take multiple images to make a panorama or HDR. D800 makes this easier, that much is clear.

The more I focus on gear, the more I drift away from the subject. It is in my nature to dive into the details, but I feel that my performance suffered in the past. I think it is important to know your equipment, but essential to let go of the technical details and focus on the intended results when it comes to the making of the image.

It all comes down to the optics in the future, I feel that Nikon has a edge in the wide and Canon in the longer end. As a photographer, I use the longer optics more and it makes sense for me to stay with Canon.

Canon had been ahead for a while but with Nikon catching up fast in recent years. Right now it looks like Nikon has the edge, I am curious to see the development in the future. I celebrate the battle and enjoy the benefits.

[MrSmith]

the reason i thought the MkIII had less noise at 160/320/640 was previous experience with the II where the files are cleaner at these iso's and i only ever move between these iso's so going from 160-400 would be worse than going 160-640.
and from this graph http://home.comcast.net/~NikonD70/Charts/RN_ADU.htm#EOS%205D%20Mark%20III_14

maybe i am interpreting the graph wrongly (i'm a humble photographer not an academic) but the results of using the 5dII daily for the last few years have shown better results at these iso's. i believe the moving image MkII users have formed similar conclusions

[Keith Reeder]

Pretty much expected, really. The 5D3 was never going to compete on RAW image quality.

Simply not true. What might be true is to say that the 5D MK III was never going to compete in DxO's game.

But the fact remains - and fact is the word - that the results people are getting from the Canon are fantastic, regardless of what that bloody website might have to say about it...

[shadowblade]

Simply not true. What might be true is to say that the 5D MK III was never going to compete in DxO's game.

But the fact remains - and fact is the word - that the results people are getting from the Canon are fantastic, regardless of what that bloody website might have to say about it...

Seen this? http://www.fredmiranda.com/5DIII-D800/index_controlled-tests.html

The DxOMark scores are reflected in reality. Compared to the D800, the 5D3, particularly in the shadows, is a noisy piece of crap.