I recently had the opportunity to shoot with a Canon EOS 5Ds for a couple of days. The Canon EOS 5Ds is the highest pixel density full frame DSLR currently available, with a 50MP CMOS sensor. That's about 40% more pixels than the 36MP Nikon D800/D810 and more than double the pixels of the Canon EOS 5D MkIII. I should note here that the EOS 5Ds I had a chance to use was a pre-production sample from Canon. It's possible that slight changes might have been made to the software that the production models use, but I believe that the results I got are pretty representative of what you'd see from a production camera.
In principle, the more pixels you put on a full frame sensor, the higher the image resolution. The downside is that more pixels mean smaller pixels and smaller pixels mean higher noise and lower dynamic range. The question then becomes whether the advantages of increased resolution outweigh any increased noise and lowered dynamic range. I'll cut to the chase here and give you the answer upfront. With the Canon EOS 5Ds they certainly do. Resolution is significantly increased, while noise remains low and dynamic range stays good.
I won't go into great detail describing all the controls and features of the EOS 5Ds because for the most part they are very similar to those of the Canon EOS 5D MkIII. Size shape and control layout are all pretty much the same.
The top deck has the usual layout with an illuminated LCD screen and buttons for direct access to functions such as white balance, metering pattern, ISO setting, flash exposure compensation, drive mode and AF mode.
The rear of the camera has a fixed LCD screen, the QCD, the multi-axis joystick and buttons for menu selection, playback, screen display etc. All pretty conventional EOS 5D MkIII stuff.
|Canon EOS 5Ds||Canon EOS 5D MkIII|
|CPU||Digic 6||Digic 5+|
|ISO range||100-6400 (50-12800 extended)||100-25600 (50-102400 extended)|
|Max Frame rate||5fps||6fps|
CF UDMA 7
|Metering||150K pixels (IR+RGB)||63 zones (RG/GB)|
|Umcompressed HDMI out||No||Yes|
|AF zones||61 zones, 41 cross, 5 double-cross||61 zones, 41 cross, 5 double-cross|
|USB||USB 3.0||USB 2.0|
As you can see, the major differences here are the pixel count and the maximum ISO settings. The EOS 5D MkIII is also better for pro video use since it can output uncompressed HDMI video and has a headphone jack for monitoring sound during recording. Metering under very difficult conditions may be better on the 5Ds since it has the same flicker detection as the EOS 7D MkII, which warns of lighting flicker (e.g. form some fluorescent tubes) and can synchronize the shutter in continuous mode so that it only fires when the light is at a peak. This ensures the same exposure for each frame. The metering system is also more sophisticated than that of the 5D MkIII in that it can analyze for both RGB and IR and can even pass along information to the AF tracking system to better allow AF to follow a moving subject. If you shoot JPEG bursts of more than 500 images or so the EOS 5D MkIII may be better, but few photographers will ever want to do that.
The EOS 5Ds also has a new feature which allows a short, user selectable, delay between the mirror going up and the shutter opening, or rather the exposure starting. The shutter still opens as the mirror lifts, but the exposure is electronically delayed by 1/8-2 seconds.
Here's where we get down to the major difference between the EOS 5Ds and Canon's other full frame DSLRs. The comparisons here are between the EOS 5Ds and an EOS 6D because those are the two full frame cameras I had available for the short time I had with the 5Ds.
First let's take a look at the difference when using a very good lens, the Canon EF 85/1.2L, stopped down to f4 for optimum image quality.
Here's the full frame (same for both the 5Ds and the 6D):
Now looking at the center of the image shot with the EOS 5Ds as a 100% crop:
And now here's the 6D image, scaled to the same size as the 5Ds image:
There's a very clear and obvious difference when the images are looked at at this scale. The 5Ds is by far the sharper image. But does this only apply when using a very expensive and very sharp "L" series lens stopped down to its optimum aperture and looking at the center of the image? Would the resolution advantage be lost when using a "lesser" lens? Well, the most "lesser" lens I own is probably the plastic mount EF 22-55/4-5.6 lens originally designed as a kit lens for the EOS iX Lite APS film camera back in the late 1990s.
Here are crops taken from images shot with this lens wide open at 55mm. First using the EOS 5Ds:
And second, a crop from a similar shot taken with the EOS 6D, scaled to match the 5Ds image:
Even with a cheap consumer kit lens from 15 years ago, the image from the 5Ds is still much better, showing that the resolution advantage of the 5Ds should be maintained with just about any lens you care to put on it. I suppose if you could find a REALLY bad lens the difference might start to become less obvious, but nobody interested in resolution is going to be using such a lens!
I also looked at the edges of the image which were softer than the center. Even there the EOS 5Ds clearly showed more detail.
For some reason which Canon have not yet revealed, the EOS 5Ds is limited to ISO 6400 (12,800 with ISO expansion enabled). The Canon EOS 7D MkII, which has very similarly sized pixels and so should have similar levels of noise, has native ISO settings up to 16000 (with expansion to 51200). So is noise at ISO 6400 on the EOS 5Ds the limit of acceptability? The 100% image crop below, shot at ISO 6400, should answer that:
There's no excessive noise here, even in the areas of uniform tone where it would be expected to show up most. In fact the image quality at ISO 6400 is good and there would seem to be no technical reason to cap the ISO there. Perhaps it's a marketing decision to further differentiate the 5Ds and 5D MkIII?
The EOS 5Ds breaks no new ground here. Canon are obviously using the same chip architecture they have been using for all the Canon sensors. This gives good, but not remarkable dynamic range. The Nikon/Sony sensors do better at low ISO settings, probably because they use a different sensor architecture with the A/D converters located on the sensor chip itself rather than being external. While I didn't attempt an accurate measurement of DR, I did shoot a series of images at ISO 100, each one progressively more underexposed, then I compensated the RAW files to see how much noise developed in the severely underexposed shadows:
As you can see, you can pull back detail without inducing excessive noise when the image is 3 stops underexposed, but by 5 stops a significant amount of noise is creeping in. This is pretty consistent with a camera like the EOS 7D MkII. It's probably a dynamic range of around 12 stops, while the Nikon/Sony sensors would show about 14 stops at ISO 100.
It should be said that the higher resolution of the EOS 5Ds images will only really become apparent to most people if you either heavily crop the images or you make large prints and view them from a close distance.
I did some tests, looking at prints form the 6D and the 5Ds. At 20"x30", viewed from a distance of 3ft, there really wasn't much difference, though approaching to a distance of 1ft revealed that the 5Ds print was sharper. It's hard to give exact numbers as to how big a print you need to see the difference because it depends on the viewing distance and how good your eyesight is, but it's certainly true that for smaller prints the difference will be small.
One thing the extra pixels do give you is the ability to crop. If you crop the 5Ds image down to APS-C size you end up with 19.6MP, about the same as the current generation of Canon APS-C DSLRs, and you end up with similar image quality too. So if in addition to your EOS 5D MkIII full frame DSLR you were carrying, say, an EOS 70D APS-C DSLR to take advantage of the "focal length multiplier" when shooting with long telephoto lenses, with the 5Ds you wouldn't need the extra APS-C DSLR since a cropped image would be just as good.
Here's a look at another image that shows what the EOS 5Ds can do. First the whole frame shot with an EF 85/1.2L at f4
And now just a 100% crop from near the center:
The EOS 5Ds has a sibling, the (which costs $200 more). It's the same camera with the same sensor, but the effect of the anti-aliasing filter over the sensor has been cancelled out. This give even higher resolution, but at the cost of possible increased moire fringing and false colors in areas of uniform fine patterning. The problems can be somewhat reduced in software, but that needs extra post-exposure processing to get the best results. I haven't yet had a chance to shoot with the EOS 5Ds R, but reports from Nikon users shooting with cameras that use a sensor without the effect of a low pass anti-aliasing filter suggest that the moire and false color issues aren't a big deal 99.9% of the time.
The lives up to its promise. The 50MP high resolution sensor gives images of outstanding resolution, rivalling those from medium format digital cameras and exceeding those of any other 35mm DSLR. The overall camera performance is excellent, noise is well controlled and dynamic range does not seem to have suffered in comparison to the and . It really doesn't have too many downsides, but it has some, such as a street price that's $1200 more than the EOS 5D MkIII at the time of writing (07/15). The ISO has been limited to ISO6400 for some reason and Canon didn't include provision for headphones or uncompressed HDMI streaming which may be of concern to video shooters.
Ignoring the $1200 price difference, I can't see too many applications in which the EOS 5D MkIII would be the preferred camera. Perhaps photojournalists, sports and nature shooters would chose it for the expanded ISO range, slightly faster frame rate and larger image buffer and serious video shooters would choose it for the uncompressed HDMI output and headphone socket. However for landscape photographers, portrait photographers and commercial product photographers the EOS 5Ds would seem to be the more desirable camera.