Is the 5D MKlll better at high ISO than the D800

Discussion in 'Canon EOS' started by jrcrowe, Mar 16, 2012.

  1. I was looking through the Nikon forum and they were talking about comparison photos between the D4 and D800. You can also compare them to the 5D Mklll here The results are surprising!
     
  2. Unless you intend to shoot only JPEG images, it is more useful to look at RAW images. Not sure why Canon cannot match the RAW processing qualities of their Digital Photo Professional software in-camera. Canon's own official sample JPEGs look awful even at low ISO but a RAW file I downloaded from Imaging Resource looked great.
     
  3. What do you find surprising John?
     
  4. I really expected to see much less noise in the D800 at ISO 6400 and above but the 5D Mk lll seems to outperform it. I had seen some photos from the D800 at WPPI, there were some night shots displayed that were quite sharp at more than 100% with little noise, and some high ISO shots that looked very good. I am surprised that the 5D Mk lll seems to outperform it in the test shots. Many people were disappointed at the specs and the price of the 5D Mklll, me among them. Its only preliminary but it looks like the two cameras may be more equal than the numbers suggest. The 5D Mk lll may even be worth the higher price.
     
  5. Well the D800 crops at 100% are considerably more enlarged, and so is their noise, than the 5D MkIII, also, despite assertions and theories to the contrary it does seem to be that for same generation sensors denser pixels in the same area do not perform as well at high iso than their less dense cousins, even when you resample down to match resolution.
    Both cameras will take superb images, but realizing the full capabilities of resolution from either will be down to very good technique. Bench tests like those linked to will always perform well, but those situations are a world away from real world shooting and I think many will be disappointed in their own results.
     
  6. despite assertions and theories to the contrary it does seem to be that for same generation sensors denser pixels in the same area do not perform as well at high iso than their less dense cousins, even when you resample down to match resolution.​
    Seriously, Scott?
    There are no "same generation" examples to prove that. None whatsoever.
    The notion that the 5D Mk III and D800 sensors are "same generation" simply because they've appeared at about the same time, is just revisionism to support your baseless assertion.
     
  7. Well if you just use the words in their correct context and don't just fly off the handle at assertions you might not agree with.
    By definition the 5D MkIII and the D800 are the same generation, that is the meaning of the word! Trying to assert that they are not in the same generation is revisionism.
    Now, Keith, show me a same generation denser sensor, of the same design, that performs better at high iso than a lower density one.
     
  8. Forgetting about the comparison, the Canon looks sweet at high ISO. The ISO 3200 and 6400 are excellent. 3200 is really clean and smooth.
     
  9. Apples and oranges. These two different cameras aren't trying to be each other. Yeesh.
     
  10. Why are simpleton bullies like Keith tolerated on photo.net? I guess it's because we tolerate the intolerant. Good on us for that, I suppose, though I feel sorry for those naive members who actually believe everything they read on this site.
    Do you know what the words "same" and "generation" mean, Keith? And are you able to put those two meanings together? Does the word "contemporaneous" ring any bells? Or are you, as your words seem to imply, either living in another universe, or brain dead?
    I'm sorry for the stridency of my remarks, but I believe that it's incumbent on us to call out the bullies in our midst.
     
  11. The D800 is 36MP vs 22MP of the 5D III. I would expect the 5D III to be the better high ISO performer. No surprise for me. I'd expect the D800 to be only slightly better at high ISO than the 7D, given the pixel density, which doesn't really impress me that much. To me, it seems a 36MP full frame camera is geared towards landscape shooters that use ISO 100 and a tripod, not a shooter looking for low light performance.
     
  12. Matt,
    I disagree, they are two manufacturers products aimed at the same market segment. They are both high end "prosumer" models, both full frame with very similar specs and prices and released at the same time. They are in direct competition for a particular market segment, just like the D700 and 5D MkII were, the previous models had bigger feature differentiation but they were still in direct competition.
     
  13. I am not surprised. It is still easier to get better high ISO performance from less dense sensors.
    These cameras are not trying to be each other, however, in my opinion, Nikon and Canon did try to be more like the other and actually thought the cameras they were developing would be similar. Each overshot the other. Canon wanted a better high ISO camera to compete with the D800 that they were expecting to see (a marginally higher res D700 with better high ISO) and Nikon wanted a better high res camera to compete with the 5D III that they were expecting to see ( a 5D II with 30+ megapixels).
    Funny how it turned out, and judging by people here on photo.net it seems that both Nikon and Canon should be apologizing to their customers for their misjudgements. Nikon has already backtracked by saying that the D800 is not meant as a replacement for the D700. Canon is sticking to their guns that better high ISO, marginally faster fps, and better AF, is worth $1000 to $2000 over the 5D II. Yes, I just noticed you can buy a refurbished 5D II, which will have similar low ISO performance, directly from Canon USA for less than half the price of a 5D III.
    P.S. I am not the original poster.
     
  14. Well, if you want to compare "same generation" cameras with different pixel density, you should compare 1Dx to 5D III.
    Guess which will have better high ISO performance?
     
  15. I'd expect the D800 to be only slightly better at high ISO than the 7D, given the pixel density, which doesn't really impress me that much.​
    It's dangerous to make an extrapolation based upon pixel density alone. We don't know what kind of firmware processing is built into any of these new sensor chips or how efficient they are at collecting photons.
    From all reports the D800 exceeds the High ISO performance of the D700. I fear that the 7D, which does a remarkable job for having a lot of pixels on a small sensor, isn't going to be in the same league with either of these cameras.
    Full disclosure: I own a 7D (which I don't plan to sell anytime soon), I used to own a D700, and I have a D800E on order. I have no brand-based bias in this discussion.
     
  16. At first glance, the Canon does seem to offer a cleaner image than the Nikon, but I'm honestly not too sure exactly what I'm comparing. When I compare the 5DIII to itself at different ISO settings, progressively higher ISOs show progressively more aggressive noise smoothing, to the point that considerable detail and contrast are lost, so there appears to be some dynamic process afoot. I'm not sure I consider this "good." I'm wondering how the images would look with noise left intact, sparing sharpness and contrast.
    This may be old news to many people, but I admit it's not what I expected. I presume these are in-camera jpegs. Would/could DPP do the same conversion? I suppose we don't know yet how the RAW files compare at different ISOs. Or do we?
     
  17. Internet IQ comparison is always difficult although I would expect the 5DIII to outperform the D800 at higher ISO. This is
    really just physics as both cameras have the same generation of signal processing and sensor. While the engineers
    differ and the algorithms also differ the larger Canon photo sites will give a better Signal to noise ratio ((Eb/No for the
    pedantic)
     
  18. Notwithstanding the limitations of comparing JPEGS, I have to say that I would be pleased to shoot either of them. The Canon looks very clean at high ISO, and it is not so far behind in resolution even at ISO 50--where both are very clean. That should not be too surprising, though. A camera with twice the resolution of the 5D III would not have 44 mp, but 88. (Resolution is a linear function, being defined in terms of line pairs per mm.) Comparing the square roots of 22 and 36, on the other hand, indicates that even on purely mathematical grounds alone, the Canon should have about three quarters the resolution of the Nikon, not 22/36, which at 11/18 would seem to imply (falsely) that the Nikon would be almost twice as good in terms of resolution as the Canon. (I realize that I presented all that in a confusing way, but those who understand resolution will, I hope, get the point.)
    Given everything else that the Canon comes with, I would say that the advantage seems to go to Canon. The resolution is not equal to that of the Nikon, but it is closer to it than one might imagine just by comparing the total number of megapixels of both.
    I used to say that I would not ever need more megapixels than the 5D II can give. Well, now we have that (and a tiny bit more in terms of resolution), but with a lot of other improvements. Right now I am happy that went with Canon back there with the original 5D in 2006, when the price had already slipped almost a thousand dollars. When it does that for the 5D III (adjusted for inflation), then I hope that I can buy the 5D III. I do not think that I will be trying to buy the Nikon. The marginal value of the added resolution is simply not that great, and on everything else the Canon is clearly superior.
    --Lannie
     
  19. I just put the information here about latest comparison result in Japan between D800 and 5Dmk3.
    ISO: D800>5Dmk3
    Dynamic range: D800>5Dmk3
    Resolution: D800>>5Dmk3
    Price: D800>>5Dmk3 (D800 is amazingly cheaper than 5Dmk3.)
    Canon users look doing funeral ceremony against Nikon users Rio's Carnival.
    Also, D800 and D800E's order percentage is 46% vs 54%.
    And waiting time from order to in coming is now 2 months for D800, 5Dmk3 is not informed yet.
     
  20. I think you messed up your ">" signs.
     
  21. Internet IQ comparison is always difficult although I would expect the 5DIII to outperform the D800 at higher ISO. This is really just physics as both cameras have the same generation of signal processing and sensor.​
    The physics says nothing of the sort.

    The physics says whether a sensor has one pixel, 4 pixels or 40 million pixels, its overall light-gathering capacity will be the same.
    The physics says it's the size of the "window", not how many panes of glass it's made up from, that controls how much light it will collect.
     
  22. That might be what the theoretical physics says, the actual engineered sensor doesnt work at the theoretical level though, for many reasons, and empirical results just don't support your beliefs, there are countless examples of higher density pixel sensors, even when they are
    younger, not performing as well at high iso (or at the pixel level at base iso) as lower density sensors. I have posted my
    own test examples many times that illustrate that.

    Whether the increased noise is too much to affect the image, or not be process-able away or any other symantec game,
    doesn't alter the fact that countless images readily available on the Internet, from reputable sites, all support my initial
    assertion and nobody has provided a single image to rebuff it.
     
  23. I'm wondering how the images would look with noise left intact, sparing sharpness and contrast.
    This may be old news to many people, but I admit it's not what I expected. I presume these are in-camera jpegs. Would/could DPP do the same conversion? I suppose we don't know yet how the RAW files compare at different ISOs. Or do we?​
    Right, what really matters isn't what comes straight out of the camera but what can be done with it in PP. A significant scrambling factor in the mix are/will be performance of different existing & future versions of the various third-party RAW converters with files from either camera, or even the continuously improving performance of noise suppression algorithms for jpeg files. Not to mention the times needed to PP 24 Mpx versus 36 Mpx files on everyone's slightly to greatly different hardware configurations.
    And I'm not familiar with the details of i-r's test procedure, but presumably both cameras were shot at +/- zero Ev. Do we know whether their iso scales are truly matched, or do one or both benefit more substantially from exposing to the right?
     
  24. As to the irrefutably definite certainties of physics, that reminds me of Click & Clack aka Ray & Tom Magliozzi's "Understanding Engineers" ;)
     
  25. I suppose we don't know yet how the RAW files compare at different ISOs. Or do we?​
    Imaging Resource has posted both the D 800's NEF files and the 5D Mark III's CR2 Raw files. I downloaded them yesterday. Once I update my ACR to version 6.6 I will give them a test.
    http://www.imaging-resource.com/
     
  26. @Keith
    The total light gathering abillity is of course the same - but it is completely irrelevant for high ISO perfomance. What matters is the `signal to noise ratio' which depends on the pixel size (i.e. exactly how many panes of glass you have)
     
  27. What matters is the `signal to noise ratio' which depends on the pixel size (i.e. exactly how many panes of glass you have)​
    Even number & size of pixels (panes of glass) are not perfectly correlated as the gaps & edges between them vary from one sensor design to another. Then there's differences in front- versus back-illuniation designs, the various filters that cover the sensor itself (bayer, anti-aliasing, IR/low-pass etc) etc etc
     
  28. Romauld,
    "The total light gathering abillity is of course the same"
    It would only be the same, in theory, if there were no gaps between the pixels, more pixels demands more light losing gaps, and so far micro lenses have not been able to overcome that completely. There would also have to be no interference between pixels, even in lab conditions with super cooled sensors this is not possible. Add in your point about S/N ratios and it is amazing that sensors like the 7D perform as well as they do, but it doesn't alter the fact that even at base iso a higher density sensor produces more noise than a lower density one. The theory is one thing, the engineered application something quite different.
     
  29. Why are simpleton bullies like Keith tolerated on photo.net?​
    I am no fan of strident remarks by anyone (everyone could perhaps benefit from some relaxation techniques). However, keep in mind that you have resorted to name calling, and Keith has not.
    Just sayin'.
     
  30. IMO, despite the price similarities, and similar position within their manufacturer's 'spectrum' , these two cameras are 'aimed' at completely different market segments.
    It is my belief that the D800 is 'aimed' at studio professionals, and those who need to maximize their output resolution (for, say, billboards). The resolution clearly puts the camera into what had previously been MF range of output, with many of the conveniences of DSLRs (including lens range, size, flexibility, speed, etc.). The 800E even adds moire and false color noise to add critical sharpness.
    OTOH, the 5D3 seems aimed at pros working outside the studio, since, despite it's lack of increased MP resolution, it has significantly improved it's AF system (almost a top of the line 1series system), it's burst rate (6FPS), and it's ISO range. All those features are utterly pointless inside the studio environment, where you have complete control over lighting, subject, time, etc.
    So, yeah, no surprise, HIGH ISO noise is better from the 5D3 than from the D800. Lower than ISO6400, you'll have a hard time telling the difference, but by 12800, there appears to be a significant difference between the two - that is, the 5D3 seems significantly better than the D800 @ ISO12800+ in regards to noise.
    Of course that's not the whole story.
     
  31. The physics says nothing of the sort.

    The physics says whether a sensor has one pixel, 4 pixels or 40 million pixels, its overall light-gathering capacity will be the same.
    The physics says it's the size of the "window", not how many panes of glass it's made up from, that controls how much light it will collect.​
    Keith, I believe when he said physics he was referring to the ELECTRONICS of the sensor. The problem is if you make electronics smaller and smaller they are more prone to noise. I heard the space industry does not use Intel i7s in there space vehicles and probes. One reason is it takes a long time to thoroughly test and vet new processors. The second is processors in space are subjected to all kinds of charged particles and the smaller circuits are more prone to errors (noise) when struck by these particles. A 486 chip has bigger more noise persistent circuits.
    As people cram more and more photo sites onto the same size real estate the electronic sensors have to get smaller and thus by the rules of physics more prone to noise. As you pointed out the sensors receive the same amount of light, but all things being equal the one with smaller photo sites will have increased noise. With each generation of denser photo sites chip manufactures also have to implement new techniques to reduce noise. What a lot of use want to see maybe a 12 megapixel FF sensor with modern noise technology. And I don't mean the crappy just smooth everything out (including detail) technology. I'm speculating that if they dropped the pixel count to something reasonable they could back a pretty sick low light performer. I am not an engineer so I could be wrong.
     
  32. Hi, John,
    I use the sigh ">" as "better than" meaning. Not straight as "greater than" of algebra.
    I'm understanding the correct use but this sigh is used quite frequently in general purpose in Japan. Sorry that.
     
  33. Well, I think they are aimed at exactly the same audience, and I think John Crowe may be right: both manufacturers sampled their customers and professional users and produced what their surveys suggested they should do- the result was cameras with different capabilities for the same audience. I think the dissatisfaction from both sides with both cameras comes from the fact that people are not 100% honest when answering questionnaires. This is a well known feature of all polls and surveys. Of course, I'm guessing about this - like everyone else.
     
  34. I suppose 'consumer surveys' suffer from this, but surveying professionals is a completely different ballgame. When your output is based on the capabilities of the tools you have at your disposal, you take very seriously questions about product improvements. Especially if you have strong brand loyalty (which, by necesity, most pro photogs do).
    That said, unless Canon comes out w/ a comparable camera to the D800 (which the 5D3 is not) soon (or announces it soon anyway), this FY (probably in Q3 or 4), instead of replacing my 5D2s w/ 5D3s, I'll probably keep the 5D2s and add a D800E + some Nikor glass for my limited studio, and portraiture work. Why would I take this unprecedented step (for me anyway ;-) )? Because it seems the D800E meets my needs far better than any current or announced Canon product.
     
  35. This sounds a bit like the story of blind men trying to describe an elephant. If the pane is the pixel, the size of the pane is what matters. That notwithstanding, what we do with the light that falls on each pane is what makes up the image. The only real questions are whether the tinier pixels in the D800 perform on par with the fatter Canon pixels in low light; and if not whether the differences are meaningful.
    Making a same quality pixel with 28% less light would represent a huge leap in CMOS light sensing technology. Without the corresponding advance in sensor technology, 28% less light gathering area equates to 28% useful light, somewhat less than a half f-stop. Is this 1/2 stop a meaningful difference? Rather than shoot at ISO 3200, the D800 shooter will have to limit herself to ISO 2500. The difference is exposing for 13 seconds rather than 10 seconds, for example. This seems a pretty small difference for the heated debate.
    Putting the pixel dimensions in more tangible terms, printing the D800's 7200 pixel wide image at 240 dpi yields a 30" wide image, compared to 24" for the 5D3's 5700 pixels. Equivalently, at 120 dpi, 60" compared to 48". That also seems a pretty large difference, but can you really tell the difference between 95 dpi and 120 dpi? I think maybe I can, but then, maybe not easily.
    The difference in pixel count can also be equated to a zoom crop factor. Inside every D800 shot is an equivalent 5D3 shot, albeit (presumably) with some ISO speed penalty. Cropping a 14mm D800 shot yields an 18mm focal length field of view 5D3 equivalent of identical quality (all else held equal). This comparison grabs my attention for some reason. It means 6" of blank white border on a 60" wide print that the other camera fills.
    Looked at individually or in aggregate, the differences are huge, and at the same time, each borders on insignificant or barely noticeable. It does seem a paradox, if we look only at 36 MP versus 22 MP.
     
  36. Michael,

    That is a complete load of irrelevant, made up, rubbish, again.
     
  37. Try to keep things civil guys. There are a few comments in this thread that border on the unacceptable and if any more are posted the thread will be closed.
    Feel free to agree to disagree, but please keep things agreeable!
    This whole discussion theoretical discussion is close to meaningless when it comes to actual practical performance differences between similar sensors. The 5D MkIII and D800 have similar sensors in that they are both Bayer Matrix, microlensed, front side illuminated CMOS devices. You can certainly say that ALL ELSE BEING EQUAL, larger pixels provide better signal to noise ratios and greater dynamic range. But all else is not equal and most of it is unknown. You don't know the photodiode efficiency (photon to electron conversion efficiency). You don't know the microlens array efficiency. You don't know the pixel fill factor. You don't know the Bayer matrix filter characteristics. You don't know the amount of electronic noise added by the readout electronics and A/D conversion electronics and amplifiers. You could make an argument based on shot noise (which is the same for all since it depends only on photon arrival statistics), but pretty much everything else is a variable that you (not being a Nikon or Canon engineer) don't know. Shot noise is dominant in the brighter parts of the image. Johnson–Nyquist thermal noise is the main contributor to dark noise. You also don't know the nature and effectiveness of any on-chip hardware noise reduction circuitry associated with the sensor. All of these things affect the RAW files. When you look at JPEGs you then have to add in whatever processing is done to the original data.
    What matters, and what actually tells the story, are comparisons and analysis of images taken with typical production cameras under standardized conditions.
    So most of the arguments made on the basis of theory in the face of unknown facts are noise...
     
  38. I'll probably keep the 5D2s and add a D800E + some Nikor glass for my limited studio, and portraiture work. Why would I take this unprecedented step (for me anyway ;-) )? Because it seems the D800E meets my needs far better than any current or announced Canon product.​
    I'm curious. What exactly are you doing that requires more than 22 MP? Or more importantly what are you doing that is so time sensitive that you have to spend money on lenses for a different camera system instead of just waiting for the inevitable 36 MP offering from Canon? I guess you said "needs" plural so maybe there is something other than resolution and noise that you are looking at.
    I have Canon gear but if a Nikon digital shooter with an extensive setup asked me whether they should switch because of a single camera body I would probably 99.9% of the time say no. Digital bodies come can go. I guess some people think there are some better offerings in the Nikon wide angle arena. Some people have also said the Nikon flash system is better. I don't know how true either of those statements are but if they are true and those are something people need I guess switching would be worth it.
     
  39. The same argument was done betweem 5dmk2 and D700. The result was 5dmark2 won in some fields.
    Simply same thing happened between D800 and 5Dmark3. And both are very nice actually.
    Canon recovered the weak points of 5Dmark2 and made 5Dmark3. Nikon choosed different way to D800 with Sony sensor. Probably they opened a door to new age of camera. D800 and D800E are looked like a test machine unified with Nikon's and Sony's semiconductor technologies and patented their new optical technologies and brush upped software technologies.
    Unfortunately D800 has only 4fps since huge data transfer to the memory per a second. D800 also require high speed CPU computer and larger memory and disk storage, and quality lens for ideal photos. 5Dmark3 look better to use sport scene or animal pictures.
    Obviously, the selection is depend on users.
     
  40. It is my belief that the D800 is 'aimed' at studio professionals, and those who need to maximize their output resolution (for, say, billboards).​
    Buying a 36 megapixel camera to shoot pictures that will only be viewed from 500-2500 feet away is a gigantic waste.
     
  41. Michael Young,
    Your analysis adds some practical sense to expose the differences where they matter most.
    Thank You!
     
  42. Here are some comparisons of the 5D2 vs. the 5D3 raw files. It seems as though the 5D3 is only about 2/3 of a stop better than the 5D2 when comparing raw files. After seeing this, I ended up buying the 5D2 as low noise performance is the most important feature to me and I didn't see the 5D3 being worth the extra cost for not even a full stop.
    http://forums.dpreview.com/forums/readflat.asp?forum=1032&thread=40829181
     
  43. It makes little to no difference.

    For the most part, whichever cam you get, in the end how your photos speak will be pretty much the
    same.
     
  44. surveying professionals is a completely different ballgame.​
    This is a general characteristic of people, in my experience, professional or not. Professionals think they know what they want, but then when they see something new they want that too. In psychology it is called incentive relativity - most people compare themselves to others not to objective needs. You can see it in action in all of these threads: MkIII vs D800, or, reversed, a few years back in MkII vs D700 discussions.
     
  45. Scott,
    If you can't tie a number to it, the minutiae is still every bit as meaningless as the lite beer tastes-better/less-filling "debate". Once you can say with some certainty that one camera is between 1/3 to 2/3 stops better in some or all qualities, you can start to put some value on the differences. I don't know which you find to be rubbish, the concept, or the numbers.
     
  46. +1 for Bob. Let's wait until regular production units are available and tested. The whole discussion is meaningless without tests being carried out in controlled circumstances.
     
  47. A 36ft x 24ft billboard viewed from 100ft away needs the same resolution as a 3x4.5" print viewed from a distance of a foot.
    Even if we had precise numbers on production cameras, it's not going to make a damn bit of difference to 99% of the photographers who use them. It's like choosing between a Ferrari and a Porche. Does it really make any difference if one does 0-60 in 3.5 seconds the the other takes 3.7, or that one tops out at 168mph and the other goes on to 178mph? Both are much, much more capable than 95% of the people who drive them and neither is likely to be pushed anywhere near their limits by their owners.
    Give me a Nikon D800 or an EOS 5D MkIII. I'll be more than delighted with either one as a present. I don't need to see detailed image comparisons. I know they'll both be good enough for any use I could put them to. If you're going to buy one, neither is good enough or game breaking enough to dump a bag full of lenses and switch (or duplicate) systems.
     
  48. FWIW, here are 2 RAW comparisons of 5DIII vs. D800 (NR0) downsized (bicubic) to 5DIII file size (ISO 25,600). Please view images at 100% by following the link.

    5DIII (top) | D800 (bottom)
    Link to full-size image


    5DIII (top) | D800 (bottom)
    Link to full-size image
    I downsize the D800 to 5DIII sizes because when we're talking about *noise*, usually I'm shooting at high ISO/in the dark, so that's probably not a landscape, so I probably won't care about 36MP (there are a host of other reasons why I think the downsizing of the camera with the disadvantage in terms of pixel size, but advantage in terms of resolution, is most relevant... but I won't get into them here).
    So when we look at the RAW (processed in ACR 6.7 using identical settings), the cameras appear to perform rather similarly in this scenario. To my eye, it's a bit of a wash. To some other folk who've seen this comparison, some claim the 5DIII has a slight advantage. Any small advantage may be magnified in actual shooting scenarios, which may be much darker than this controlled setup.
    So take what you will from this comparison.
    Excellent points have been made re: all the uncontrolled variables that deter one from making general statements about one sensor vs. another. I would argue that the D4 vs D800 would be a valid comparison of same-generation sensors. In that case, each pixel of the D4 will have higher SNR compared to D800, and so shot noise will be reduced for the D4 pixels. Shot noise of D800 can be decreased by downsizing D800 image to D4 ('software binning'), but this will not perform as well as 'hardware binning' or actually using the D4 sensor since the D800 will have much more read noise injected into the image b/c more pixels are read (if anyone can comment about how read noise is reduced, i.e. by what factor mathematically?, during 'software binning', I'd be much obliged!).
    But given that Nikon sensors seem to have better read noise than Canon sensors (sensorgen.info), I wouldn't be surprised if the similar performance we see above despite Canon's higher SNR per pixel is b/c of better read noise on the D800 (I'm guessing here).
    Also, decreased read noise drastically helps dynamic range (recoverable shadow detail), and preliminary calculations from some show that DR is not improved on the 5DIII vs. the 5DII. That would hint at the D800 having better DR, which'd just be crazy given that its pixels are smaller. The method for those DR calculations were to take an underexposed black frame & an overexposed white frame, then apply the following formula:
    DR = log[base 2] (max pixel signal in white file/stdev of pixels in black file)
    I did this myself with 5DIII black/white RAWs using IRIS & got ~11.2 stops. But I don't know if this methodology is valid because I feel like post-capture/pre-RAW-file-writing processing could muck with the stdev of pixels in the black file. Skewing this calculation. For example, the black signal in 5DIII has a mean of 2048; black pixels from D7000 has a mean of 0.42... so of course the D7000 will have smaller stdev around a signal of 0.42! But, these numbers match up with DXOMark numbers, so...
    Thoughts?
    Anyway, I anxiously await real world tests!
    Cheers,
    Rishi
     
  49. I'm taking this thread a bit late, but here's my own input on this issue.
    I more or less did what Rishi did, with the same raw files, downsampling the D800 files to the 5DIII files.
    At 100%, I could see a very slight DR advantage to the Canon : a minute additional amount of detail in the dark cloth, and at higher iso, some very slightly better tonal transitions. It is my opinion that these differences are so marginal that they are of absolutely no consequence.
    At full res, there is a bit more noise in the D800, but when I compare the files in more realistic conditions, displaying them at res emulating a 30x40cm or 40x50cm enlargement (my maximum in terms of prints), it seems to me that I would be completely unable to differentiate between the D800 and the 5D in a blindfold test, from 100 to 12800 iso, possibly even 25000iso, especially considering that a computer screen might reveal a bit more dynamic range detail than a print.
    In other words, it is my opinion that deciding on one camera vs the other strictly based on the sensor's intrinsic qualities is relevant only for photographers with the most extrem needs in terms of enlargement/ cropping (D800) or high iso (5DIII). For the vast majority of us, I would say that both cameras have a level of perfomance that far exceed our photographic needs (at least it is the case for me). It might thus make more sense to concentrate on other criteria such as, the current stable of lenses one has, respective lens offerings from both manufacturers, ergonomics, etc.
    One issue I had was my computer equipment : my aging system deals smoothly with 5D1 files or Fuji X100 files. 5DIII file conversion was a bit sluggish. D800 files conversion took anywhere between 30sec and 1mn per file. Getting the D800 for me would mean upgrading my computer equipment.
    One last thing : I'm still an avid film shooter, alongside my digital picture taking. All you need to do is shoot a couple of 1600iso color films or 3200 b/w films to see the phenomenal progress in high iso quality achieved in recent years in digital photography. It amuses me to see how people can become obsessive about marginal differences in noise levels, given that a 5DIII file at 105Kiso will have noise that is probably less conspicuous than grain on a 1600iso color film (film grain having a slight advantage in the aesthetics department, though).
     
  50. I too processed the imaging-resource raw files from the D800 and 5D3 @ISO6400. My conclusion was that noise characteristics were very very similar whether I downsized to 5D3 size or upsized to D800 size. I noticed slightly better detail retention in the texture of the cloth in the D800 files, but that is really nitpicking IMO. I think factors other than ISO and resolution will likely be more important factors when trying to make a decision between the D800 and 5D3.
     
  51. @Rishi, the 5D III is clearly the winner after down-sizing the D800 images. Do you have any comparisons made without the downsizing this is what people really want to see ?
     
  52. People keep saying the 5DIII is the winner here & I'm sorry, I just don't see it. In fact, the D800 image looks slightly less noisy to my eye. But since eye/brains are subjective, I thought I'd quantitate a portion of the two images.
    I quantitated the bottom of the black bottle between the two files & calculated a mean & standard deviation.
    Here are white lines through the regions I quantitated for the 5DIII (top) & D800 (bottom):


    Here are the results plotted in Excel (quantified in IRIS):
    [​IMG]
    Make of it what you will... but in these two quantitated areas, the pixel variation is lower for the D800 (19 for 5DIII, 16 for D800). Albeit by a very small number (3 on a 255 scale).

    I quantitated another area within this same region of the bottom of the bottle & the means came out to 16 for both 5DIII & D800, but a STDEV of 18 for the 5DIII & 15 for the D800, again giving credence to the slightly lower noise of the D800.

    Again, we're splitting hairs here. ISO performance looks virtually equivalent in this scenario.
    The advantage of the Nikon here (in terms of standard deviation of dark area) might be due to its lower read noise (Nikon sensors typically get rated as having lower read noise than Canon sensors), so maybe I should quantify a gray area as well (haha, pun intended).
    -Rishi
     
  53. When I down-sized the 5d3 and d800 raws, there was no clear winner in my opinion. I'm anxious to see the 5d3 raws from Dpreview to see if my observations stand or not.
     
  54. Do not post inline images which you have not taken yourself. Any such images posted here will be removed.
     
  55. Oops! I'm sorry Bob. I didn't know of that rule... although in retrospect, I really should have! Will definitely be careful in the future. I even forgot to credit the source... poor form. My apologies again!
    But links are OK?
    If so, then here are the lines showing the quantitated regions (5DIII on top | D800 on bottom) so there's a reference for that Excel plot:
    http://cl.ly/3C2M2c3X1Q3y3h441Z0M/QuantitatedAreas.jpg
    All images are from Imaging Resource (http://www.imaging-resource.com/).
    Cheers,
    Rishi
     
  56. Rishi, I am quite impressed with the thoroughness of your analysis. If people seem to diverge as to which camera delivers the smallest amount of noise, it is, as you say, because we have a subjective perception of it. Which area of the image we give more attention to, how we react to the shape and distribution of the noise strongly impacts the conclusion we draw.
    What your graph shows is how small the objective difference is between the D800 and the 5DIII. As objective differences decrease, subjective perception will takeover, hence the diverging opinions.
    In the end, I'll repeat what I already said. For practically everyone of us, from 100 iso to 25600, the difference between the to cameras are irrelevant from the perspective of real world photographic achievement. For someone starting a new system, factors other than sheer sensor quality should be given priority.
    Thanks again for you detailed analysis which must have been quite time consuming. The technical data you offer seems to corroborate the feeling I had when visually comparing the files in "realistic" conditions (40x50cm enlargement). This is great to know.
     
  57. Putting physics aside for a moment....... I've been using digital cameras since 1999 starting with the Kodak DC210 (1 Meg Pix), what I've seen along the way is that as you cram more and more pixels into same size area, the noise level increases. Finally someone decides they need to do something about the noise level and invest in R&D to improve things. The DC210 had ONE ISO - 140. My next camera was a DC-260 (1.5 Meg Pix) to keep the noise level down they dropped the ISO to 100.
    I had a Sony F-717 (5 meg pix), noise levels were good, ISO had a real range of 100-800, moved the F-828 (8 meg pix), same ISO range noise levels were much more pronounced. (both had the same size sensor)
    Canon actually dropped the number of pixels going from the G10 (14 meg pix) to the G11 (10 meg pix) in part due to noise issues. They kept the G12 at 10 meg pix for the same reason. Only with the G1X did they go back to 14 meg pix, but did so with a MUCH larger sized sensor
    I've seen the same kind of thing with the 5D and 5D Mark II - more noise as you cram more pixels into the same area. Did it stop me upgrading from a 5D to 5D Mark II? Absolutely not. The increase was there but not significant enough to make me change my mind. To me it's just a factor to be aware of when I'm shooting. Knowing my tool and working to accommodate it.
    Now with the 5D Mark III, it appears they have done a lot of work to improve noise levels
    Just my 2 cents based on observations over
     
  58. You have to be careful when you talk of noise increases due to higher pixel counts. While pixel-level noise increases (technically: the signal to noise ratio, SNR, is lower per pixel in a higher resolution sensor than a lower resolution sensor of equal size), it's not necessarily true that SNR is lower when the higher resolution image is downsized to the lower resolution size. Downsizing the higher resolution images increases pixel-level SNR.
    DXO's D4 vs D800 SNR comparison shows exactly this if you look at the normalized SNR curves (http://www.dxomark.com/index.php/Publications/DxOMark-Reviews/Nikon-D800-Review/Sensor-performance).
    -Rishi
     
  59. i have already taken 2x 5d3 back to the shop . despite using 2 metods to get sharper images out of it, my 5d2 takes way sharper pictures from iso 100 ++ i done ewrything i can think off but nothing helps . was hoping the 5d3 could replace both the 5d2 and my 7d .. anyone know wy this so awsome new model cant deliver sharp images ( i love ewerything else whit it) and as my 7D on a 1,6 crop it assembles the D800 on pixel density .so i dont think the d800 outperforms 5d2 or 5d3 on high iso .
    if Canon dont fix this asap,
     
  60. Statistically, the chances of having two defective bodies in a row with the same "problem" point very strongly to user error.
     

Share This Page