Image Quality

Edited from content originally published here on March 28, 2016.

Our latest test scene is designed to simulate both daylight and low-light shooting. Pressing the 'lighting' buttons at the top of the widget allows you to switch between the two. The daylight scene is shot with manually set white balance aimed at achieving neutral grays, but the camera is left in its Auto setting for the low-light tests (except Raw, which is manually corrected during conversion). We also offer three different viewing sizes: 'Full', 'Print', and 'Comp', with the latter two offering 'normalized' comparisons to more fairly compare cameras of differing resolutions by using matched viewing sizes. The 'Comp' option chooses the largest-available resolution common to the cameras being compared.

ISO performance

So what does ISO 3 million look like? See for yourself if any of the ISOs above the D4S' previous maximum ISO offer anything useful. Nikon's claims of better ISO performance due to color filter array optimizations appear to have merit: noise levels in Raw mode are slightly lower in comparison to the D4S, or any other full-frame camera including the 1D-X II, when normalized. Although the performance advantage is more obvious at higher ISOs, like 204,800, the actual benefit does appear to be minimal at best. In fact, compared to the 42MP Sony a7R II, midtone performance at the very high ISOs is fairly similar at a common viewing size, with benefits most apparent in high ISO dynamic range, or, shadow performance at in extremely low light scenes. That benefit on the high ISO end diminishes at lower ISOs: the D5 has a 2 EV deficit in base ISO dynamic range compared to the a7R II.

Although the D5 is the best full-frame performer in terms of high ISO performance, you may be wondering: 'why the lack of a drastic improvement'? At this point, we're simply running up against the best that modern silicon can do: with single electron read noise levels at the highest ISOs in some modern architectures, there's only so much performance to be gained without drastically increasing conversion efficiency or light gathering capability past the limits already imposed by the Bayer array and current (very good) microlens design.

JPEG performance

As evidenced by the gigabit ethernet port on the D5, it is designed not only to take photos quickly, but also to send them off quickly as well. Couple that with a growing tendency for news outlets to reject Raw files in favor of JPEGs from their contract shooters and the importance of a solid JPEG engine becomes even more apparent.

So how does the JPEG engine fare? In terms of detail retention, Nikon (and Canon, for that matter) have some work to do with respect to optimally balancing sharpening and noise reduction. Detail in the Raw is left on the table at both low and high ISO sensitivities, especially in comparison to Sony's more sophisticated engine. The lack of fine detail even at low ISO is likely due to large radius sharpening, which leaves behind halos at edges that aren't there in the Raw, or the Sony JPEG for that matter. That said, halos aren't as severe as with the Canon 1D-X II, which appears aggressively over-sharpened. While large radius sharpening sacrifices pixel-level detail, it does lead to higher perceived acuity at smaller viewing sizes - note how the 1D-X II appears sharpest when everything is downsized to 8MP, but at the cost of sacrificing finer detail. This sharpening is difficult to reverse if you don't want the halos or over-sharpened look, and overall we never experienced the eye-popping pixel-level detail we're used to seeing in recent Sony JPEGs at 100%.*

Colors are a strong point though, and generally appear punchier than the D5's predecessor. In fact, the D5 has some of the nicest yellows we've ever seen: Sony's yellows look positively green in comparison, and just more yellow than the 1D-X II. Nikon still renders some of the nicest greens we've seen, particularly due to their warmth. Canon still has the upper hand when it comes to reds though. Importantly, color saturation is retained at high ISOs - more so than with the D4S, and certainly more so than the Sony. By default, the D5 neutralizes warm tones under tungsten light, leading to somewhat unnatural results in low light, but you change this behavior by setting White Balance to 'Auto2: Keep warm lighting colors'.

The D5 is fairly aggressive with respect to noise reduction, but results are quite pleasing despite loss of detail. At ISO 25,600 we see smoother JPEGs compared to the D4S or the less aggressive engine of the 1D-X II, and when stretched to 51,200 or its highest native ISO of 102,400, the D5's JPEG engine really starts to come into its own. It retains color fidelity and accuracy extremely well, exhibiting a marked resistance to color bleeding and overly disruptive shadow noise. This all comes at a detail retention cost though, and the Sony a7R II more sophisticated context-sensitive noise reduction is able to preserve more detail, particularly in low contrast areas. The 1D-X II's more lackadaisical noise reduction leaves behind more noise, but without much added detail, so the D5's results are quite favorable in context.

We chose the Sony a6300 for this comparison for a more level playing field in terms of resolution, while retaining Sony's latest JPEG engine.