One of the key questions we wanted to address in this review is: how much better than full frame is the image quality from 44 x 33mm medium format. As our technical editor has pointed out, at each of the extremes where you might expect the Fujifilm to excel, there appears to be a full frame camera that can close the gap and away from those extremes, you'd expect no significant difference.

We shot a strobe-lit studio portrait and a detailed cityscape, with a couple of cameras side-by-side, so that you can see the differences. All images processed with Adobe Camera Raw, sharpening minimized and common sharpening applied in Photoshop (unsharp mask, Amount 100%, 0.6 radius).*

Studio Portrait

These images were all shot with the same lighting setup, focused as precisely as possible. The Fujifilm was set to ISO 100 and F8, as was the Canon EOS 5DS R. The Nikon D810 was pushed down to ISO 64 and F6.3, to match total light capture and depth-of-field (the higher exposure giving more light per unit area in proportion to how much smaller its sensor is, which it can tolerate thanks to a higher pixel well capacity).

The idea of this high dynamic range image is to allow an assessment of detail and tonality. The Raw files are available for download if you wish to subject them to your own processing workflow to see what they're capable of, but we're not seeing the 'grace' or 'tonality' that medium format advocates might be hoping for.

What we do see though is a slightly noisier image from the Canon, evident in the backdrop, and resulting from the lower total signal:noise ratio due to its smaller sensor (and that's before you consider the camera's lower dynamic range due its high read noise). Although subtle here, this reduces the file's flexibility if you try to increase contrast or heavily post-process your image. Interestingly, the Nikon D810 does not suffer from this problem, because its signal:noise ratio across the image is essentially the same as the GFX, thanks to its ability to capture the same amount of total light at ISO 64 as the GFX at ISO 100. This means the Nikon D810 offers base ISO dynamic range similar to the GFX (we compare it to the similar MF sensor in the Pentax 645Z here), and so its files are likely to hold up just as well to demanding post-processing, assuming you can give the camera the extra light ISO 64 requires.

Cityscape

We shot a scene near infinity with the GFX 50S, Sony a7R II, and Canon EOS 5DS R to get a look at real-world resolution differences for landscapes and distant scenes. We shot them all at equivalent focal length (more or less matched for image diagonal). Have a look below.

It would be hard to declare a winner between the GFX 50S and a7R II - please explore the scene. While the GFX generally wins, it's by an incredibly small margin, while falling behind on the left side of the frame, and pulling ahead on the right. In other words, lens decentering** may have more to do with which wins than outright system resolution between the $3800 (MSRP of a7R II + 85/1.8)*** and $9200 (MSRP of GFX 50S + 120/4) options.

Consider further that the GFX 50S has a 0.79x crop factor relative to full-frame, so the 120mm lens we shot on it is roughly equivalent to 95mm on full-frame. We shot the Sony a7R II with the FE 90mm F2.8 Macro lens, which has a measured equivalent focal length of 92mm. So, if anything, the Sony a7R II is at a slight disadvantage given the shorter equiv. focal length (you can see this in the fact that a7R II scene elements are slightly smaller).

The Canon 5DS R with a 70-200mm F2.8L IS II lens at 95mm is a more evenly matched comparison; however, it fails to keep up with the GFX 50S throughout much of the scene, albeit almost matching it in some portions of the scene. Much of this may be lens related (zoom vs. prime), but we tested the EF 100mm F2.8 Macro, the Sigma 85mm F1.4 Art, EF 70-200 F2.8L IS II, and Tamron SP 70-200 F2.8 Di VC USD lenses on the 5DS R, and while the Sigma 85/1.4 performed the best, it was too short a focal length for a 'fair' comparison (and it wasn't significantly better, either). Either the 5DS R is inherently softer, per-pixel, than the a7R II (plausible given the a7R II's shallower pixels and the 5DS R's AA+AA-canceling filter stack), or we got unlucky with 4 different lenses on the 5DS R.

Either way, the take home message is far simpler: when normalizing sharpening across cameras, the Sony a7R II achieves at worst 86% and at best 96% the linear resolution of the GFX 50S, even with a cheap, small lens, while only sacrificing approximately 1/3 - 1/2 EV base ISO dynamic range and parity in low light performance (better low light performance if you consider the faster lenses available).

That's serious food for thought.


* Some comparisons to date have showed a drastic sharpness advantage to the GFX 50S, but many of these have been due to the extra sharpening Adobe Camera Raw applies to GFX 50S Raw files relative to other cameras, something we verified in internal MTF testing. Our tests (including our studio scene) use exactly equivalent, normalized levels of sharpening, which further explains the lack of drastically improved resolution performance of the GFX 50S in our tests.

** The performance of your lens may have more to do with ultimate resolution than body itself. The flatter photodiode structure of the Sony a7R II leads to lower color crosstalk and increased MTF, while the smaller microlenses (less 'gapless') of the GFX lead to more sharpness, at the cost of some low light performance (because each microlens is throwing away some light from neighboring pixels, also potentially leading to more aliasing). In the end, it's about how much resolution you want to eke out of the system.

*** The FE 85/1.8 outperformed the 90mm Macro, but we couldn't include it because of slight focal length differences relative to the 95mm equiv. GFX 50S shot.