PetaPixel

Low Light Comparison of the Canon 5D Mark II, 5D Mark III, and Nikon D800

The folks over at NoFilmSchool recently did a low light comparison of the Canon 5D Mark II, Canon 5D Mark III, and Nikon D800. The cameras were used to film the same dark candlelit scene with the same settings, and the ISO was slowly pushed up to the cameras’ respective limits. It’s pretty striking how big of a difference in low light/high ISO quality there is in the cameras, especially in light of DxO Lab’s test results for the cameras’ sensors…

(via NoFilmSchool via Fstoppers)


 
Get the hottest photo stories delivered to your inbox.
Get a daily digest of the latest headlines:
  • Zack Griswold

    Haha…D800 sucks compared to even the Mark II. I was upset when I saw that DxO test result, but the actual results are CLEAR!

  • Sarah

    Wow there’s a really obnoxious ad from Robax on this page that pops down over the video every time you scroll over the ad. Irritating. 

  • http://twitter.com/JohnMilleker John Milleker

    Thanks, next time I buy a camera for a candle light only shoot I’ll keep this video in mind.

    C’mon people, shoot what you have. Canon people aren’t switching to Nikon and vice versa. Even if you were shooting a ‘candlelit’ scene you’d be gelling a warm light so that you don’t have to go anywhere near these high ISO’s.

    Stick with your systems, Nikon is best one season, Canon the next. It’s an endless battle. Shoot and stop worrying about these nowhere near real world tests.

  • http://www.hdcamteam.com HD Cam Team

    Bottom line (already known for lot of users): we NEVER should take DxO Lab’s test results as “real world results”.

    Those tests are VERY different than real world shooting, they are “raw” sensor tests under specific conditions.

    There are many other factors that will make HUGE impact on the final delivered image (those done by the camera processor(s) and their algorithms). This is even more evident in video/movie mode.

    Notes:

    - On the EOS 5D Mark 3, IPB still offers better image quality than ALL-I mode in many situations

    - It is possible to shoot at ISO 8,000 and 10,000 with the 5D Mark II (and other cameras).

    Just check our finding and report posted in early 2011: NON-Official ISO on Canon DSLRs

    There are some ISOs not available from the camera’s menu, yet possible via USB commands. All cameras have many surprises inside (Magic Lantern has already demonstrated it extensively!)

    Regards

  • Vashistha Pathak

     DXOMark score based on RAW sensor output only, they never publish or test the  image processor used in the system (DIGIC V and Expeed3) and  its system noise-canceling capabilities…

  • Mute

     Amen.

  • Ninpou_kobanashi

    Why are the Canon videos so dark compared to the Nikon?  Just curious.

  • Iwgl1981

    You’re right.

  • http://www.facebook.com/zerflo Er Flow

    Amen two times… 

  • Txpunisher

    This test is flawed IMO. There can be a deviation of 1/3 stop plus or minus from lens to lens. This means between any two given lenses of the same make and model can be up
    To 2/3 stops deference between them.

  • Ndt

    Do you have any more info regarding the claim that ALL-I is inferior to IPB. ALLI file sizes are 3x the size, which suggests much lower compression and better IQ.

  • http://www.hdcamteam.com HD Cam Team

    Yes, even with 3 times larger file size, sometimes ALL-I is noisier and shows more macro blocking than IPB.

    You can find many tests and comparison at Cinema5D forum. Search for posts made by member “jcs” and other members.

    Also remember the EOSHD tests at low ISO (most people focused on high ISO), they showed lot of “mosquito” / “fizzy” noise.

    Some people request ALL-I mode to be CBR instead of VBR, that could improve its performance indeed.

  • DJ Bloom

    What a horrible “test”, soft focus across all three. No testing of the lens to check if they are all the same to start with, what a disaster

  • Jack

    Looks like Nikon beats Canon on lower ISO but gains more noise and exposure at higher settings.  I prefer to stick with lower settings when possible. 

  • http://twitter.com/davidcastillo David Castillo

    I agree with Txpunisher and DJ Bloom. This test is amateur at best.

  • Ghosty Poster

    For those who bashes DxO rating, keep in mind that DxO compares picture, not video. Feel free to read a comparison on fredmiranda.com -> http://www.fredmiranda.com/5DIII-D800/index_controlled-tests.html

  • BANANAMANANAS

    Actually, I’m a Nikon shooter and am seriously considering a 5D Mark II for video since its fully compatible with Nikon lenses if you use a calibrated adapter, since focusing is manual for video anyway. I’m a student and don’t have a lot of cash, so the price difference between the D800 and a used 5D2 is important to me. Plus, it gives me the opportunity to tap into Canon lenses.

  • hfxPhoto

    DxO Labs does not rate video performance. They rate still images.

    If you are a videographer, Canon is still better at shooting candles.
    If you shoot stills…

  • http://twitter.com/cyclonetog Merv

    What?  You REALLY didn’t see how much more sensitive the D800 is, the Canons are clearly amped down to hide noise.

  • http://twitter.com/lexplex_ Lexplex

    I have a Canon 5D mkii and the ISO isn’t that good at high settings.  The vid looks under exposed.  I’m sure if I set it to the same brightness as the Nikon test at the bottom the results would be pretty similar. 

  • Joe Marine

    Only one lens was used for all three cameras – the 85mm 1.4. I completely understand what you’re saying about lenses being different, but I only own one 85mm 1.4, and so the same exact lens was put on each camera – ensuring there were no anomalies caused by the lens itself. 

  • Joe Marine

    Again, only one lens was used – and the test was conducted three times for each camera. Focus is not soft it was checked multiple times. What you’re seeing is the lower resolution of the Canon cameras in video mode. The D800 is tack sharp.

  • Joe Marine

    I’m not sure how this test could be conducted in a more professional manner – exactly one lens was used, thus ensuring that the results could not be affected by the lens itself. We pulled the lens off the camera each time we ran the test. What would you have suggested I do differently? This is the same way that anyone else would have done this exact test.

  • Joe Marine

    What if you’re shooting a documentary and you’re using one of the DSLRs because you need the low-light ability over a standard 1/3 inch camera? You’ve got on shot at a scene under almost no light – it helps to know how far you can push a camera before it breaks down. If you know you’re always going to be in no-light situations and you have to keep a low-profile, it’s clear that the 5D Mark III will be cleaner – but maybe you don’t need that ability, you’d actually prefer to have a sharper image and a clean, recordable HDMI out to go 4:2:2 to a professional codec. Then the D800 makes sense.

    This is actually a real-world test, not a chart – you can see what the noise patterns look like when the cameras are pushed to their limit. The idea is that people might be in situations where they need to know how far they can push a camera, and if they are waiting to purchase, I think it’s valuable for them to have an idea, since you’ll never see a test like this from Canon or Nikon. Yes, candlelight might be extreme, but I’ve shot in many, many extreme situations for documentary work, and I’d rather expose properly and have some noise than be completely underexposed.

  • Joe Marine

    There’s no way to underexpose anything the way the test was done, nothing was touched but the ISO. Only one lens was used for the entire test – and the F-stop stayed the same throughout, F/1.4.

  • Steve

    There’s no detail in the hair with Canon and it looks too dark.  The Nikon has lots more detail but more noise.  I don’t think it’s a useful test, as most people don’t use just one candle to light a subject.  The difference in the way the cameras expose this scene makes comparison a bit pointless.  If you were using the Canon’s, you might want to boost the exposure and that might increase the visibility of shadow noise.

  • http://twitter.com/lexplex_ Lexplex

    The exposure values are clearly different between the Nikon and Canon.  If I adjusted the exposure to give a similar histogram distribution between the two, I’d get more noise on the Canon than this video is shooting – I know this because I own one and I use it very regularly in low light situations with 1.4 and 1.2 lenses.  :-)

  • Anatole Klapouch

    This girl blinks a lot!

    Disappointed with the D800. The image is awful.

  • chasstaf

    agreed

  • http://www.fotographix.ca/ Photographer in Calgary

    Just buy them both. Simple.

  • Kelbyscott

    Get at least the histograms to look similar!!!

  • Matt

    I don’t think the hi ISO is so cut and dry when Nikon is 1.5/2stops brighter at every ISO

  • http://www.hdcamteam.com HD Cam Team

    In fact DxO rates RAW sensor performance, not even still images.

    If you shoot video Canon has advantages. If you shoot stills, it all depends on your needs. Images from 5D Mark III are cleaner from noise even at low ISO.

  • Wallerus

    D800 did really well at the lower ISO levels, but slowly got noisier and noisier. I Think all three cameras did a great job at their respective levels. As in, each one did really well at certain points. Go shoot already!

  • http://twitter.com/davidcastillo David Castillo

    So if the same lens was used for all the tests then I assume an adapter was used at some point? If so, that could also effect lens performance by 1/3 to 1 stop of light depending on the adapter used.

  • http://mike.heller.ca/ Mike

    That claim seems to be counter to a lot of the images already available on the internet.  In head to head tests, the D800 has more DR and less noise up to ISO 1600 than the 5DIII.

  • http://www.hdcamteam.com HD Cam Team

    Well, I’ve seen many tests showing different results, D800 being noisier than 5D3.

  • Charles

    That girl is so beautiful, I know this was supposed to be a test, but the music and the girl made me feel like I was watching some sort of short film. 

  • http://www.facebook.com/bradtrent Brad Trent
  • http://twitter.com/davidcastillo David Castillo

    I agree. There has to be more going on here.

  • http://www.karimsaad.net/ Karim Saad

    Maybe it is more sensitive but the noise is really annoying in my opinion and cannot be used.

  • Jens

    This test doesn’t account for the fact that Nikon cameras are at least 1/3 of a stop brighter then Canon… hence all that noise in D800 is just hidden in the canon footage… shoot with the same levels and then compare, black is a great way to hide noise!
    And who uses video straight from camera…. really… put both through some post production and see which footage has more dynamic range / noise when manipulated… what I really want to know is which camera has the more usable data to push around. 

  • http://twitter.com/LeeLovePhoto Lee Love

    What a stupid test. Come on folks NIKON or DXO never said the D800 was a high ISO camera. The highest ISO is only 6400 unless you go into the Hi1 mode. If you want to test ISO performance then the D3s or D4 are meant for high ISO usage.

    But, as a D800 owner I will tell you the D800 will SMOKE any of these cameras in dynamic range. I am amazed at the resolution and dynamic range of the D800, which is what is was designed to do.

  • http://www.hdcamteam.com HD Cam Team

    No, you can use Nikon lenses on Canon bodies with an adapter that does not affect lens performance at all.

    The adapter consist on a simple metal ring to allow mounting Nikon lenses on Canon bodies without any optical element.

    That’s an “advantage” of Canon bodies, you can use Canon or Nikon lenses (or other ones) without optical elements that may degrade the image. Of course the lens must have aperture ring to allow you change it manually, otherwise it will be always wide open.

    The only “problem” we have found (and to be considered) when using an adapter on Canon bodies is that when using lenses without any communication with the Canon body (either you use a 3rd party manual focus lens or a Nikon lens + adapter) is that Canon cameras don’t display light metering accurately.

    It “could” be the reason of the test showing Canon’s footage a bit darker than Nikon’s.

    But in fact it is NOT the reason. Why?

    Because all 3 cameras were set with the exact exposure: ISO, aperture, shutter speed and white balance) as explained on the beginning of the video.

    So, in such case (assuming there was no change of exposure level in postproduction) any difference between exposures is due to the different camera brands delivering different footage, nothing else.

  • http://www.hdcamteam.com HD Cam Team

    All 3 cameras were set with the exact exposure settings: ISO, aperture, shutter speed and white balance, as explained at the beginning of the video.

    So, in such case (assuming there was no change of exposure level in postproduction) any difference between exposures is due to the different camera brands delivering different footage, nothing else.

  • Jonas Nordlund

    Then compensate like the Canon seem to do?

    I wonder how they would have compared with similar exposures.

  • Jonas Nordlund

    That’s the big question here. Obviously the Nikon will expose noise more since, uh, the entire video is exposed much brighter and bringing out far more details.

  • Jens

    Unintuitive… but the problem is that you are using the same aperture and lens…. seems strange to say that until you realise the sensors are not the same…. you cannot assume they expose the same way…. its known that nikon exposes differently to canon in stills… 
    So in conclusion you need to use a DIFFERENT aperture on each camera to get comparative results….

    For a quick indicative look
    Push up the levels in post on the cannon footage… is it noisy?
    Drop the levels on the nikon footage… how does it compare?

    Better yet…
    reshoot a comparison… set the exposure… to a test pattern and check levels….  adjust to the same exposure…

    My guess however is that canon has more experience with processing video from the sensor… even with a retest the canon will likely best the D800.

    But no matter what early results show fortunately new firmware can do wonders… and canon can’t upgrade the sensor without shipping a new camera.. my guess time will tell how good D800 video is… 

    Finally don’t knock the DXO guys the test they do have a lot of rigour and knowledge behind then… I believe the nikon sensor is better then the canon one.

  • Three_fitty

    HD Cam if you read the DXO reports you would realise that the reason the Nikon image is lighter is because although the ISO “settings” are the same on all three model. The ACTUAL ISO being employed is different. Canon and Nikon will claim an ISO of 2000 in this case. Test will show you that it will actually be using 1800 or 1900. This is part of the marketing game. It allows them to claim super high unrealistic ISO’s. In this test Nikon’s ISO of 2000 is closer to that number than Canons. Hence the under exposure in the Canon’s. In order to match the D800 the Canon’s would have to go to a higher ISO and therefore introduce more noise. 

  • Carlo Righini

    Thanks for that response, its helpful to understand at least a possible reason for whats happening here. The issue I have is that I went into the DXO sensor data and looked at both the D800 and 5d Mk iii results. What it showed is that the D800 actually was less accurate than the Canon in terms of is actual versus stated ISO ratings. So I don’t think we have the proper answer as to why we are seeing such a big difference in the exposure values.
    http://www.dxomark.com/index.php/Cameras/Camera-Sensor-Database/Nikon/D800
    http://www.dxomark.com/index.php/Cameras/Camera-Sensor-Database/Canon/EOS-5D-Mark-III