initial review: 1/5/2009
updated: 1/13/2009
updated: 1/17/2009 see top of Performance for a long note
updated: 1/26/2009 Complete Redo

  Nikon D3x Review (Do Over)

Same body, more pixels

  Add a comment or send Thom feedback on this article.

As expected, I've received a tremendous amount of feedback on this review, both positive and negative. That includes contact from Nikon's marketing, who asked for a slight change (I had misquoted something), which I've made.

But the tale of redoing this review is more involved than just responses to readers and Nikon. It seems clear to me that the original D3x I tested was sub-par. In particular, besides the viewfinder alignment issue, there was a deeper one that I discovered very rapidly and pondered about too little when I wrote my original review. Basically, what I found is that my original D3x was noisier than it should be. The amp noise issue I originally reported, for example, was one indicator. But when I ran tests against two other D3x bodies I found that the first one I received was reporting different values for low-level noise, especially in long exposures or high ISO values.

In the interest of fairness and accuracy, I ended up returning my original body, obtained a new one, then tested it. I also briefly tested two borrowed bodies. My conclusion: my original body was the outlier in most tests in which noise measurements were involved. I'll have more to say about the issue of getting an inferior sample towards the end of the review. But the bottom line is that the other D3x bodies perform somewhat better than the first, thus I'm going to have to discount that original body as a "sample variation" problem. I've now re-run and reshot most of my tests, and am comfortable in reporting (some) new results. My overall ratings haven't changed--though I was tempted to push value to four stars for some.

You really need to re-read the entire review, as I've reworked wording almost throughout to reflect what I've found.

It's going to seem a bit surreal, but here's the bottom line: I'm (still) not able to recommend the best DSLR currently made to most of you reading this.

But, first things first: if you haven't read my D3 review, do so now. I'm not going to repeat anything I wrote there. Since the D3x differs from the D3 only in the imaging sensor and internal imaging chain, all the comments about camera features, handling, and non-image related performance such as autofocus will be the same with the D3x.

Everything I wrote about the D3 applies to the D3x, with the exception of:

Frame Rate FX 12 bits 9 fps 5 fps
Frame Rate FX 14 bits 9 fps 1.8 fps
Frame Rate DX 11 fps (with restrictions) 7 fps
FX Pixels 12mp (4256x2832) 24mp (6048x4032)
DX Pixels 5.1mp (2784x1848) 10.5mp (3968x2640)
Normal ISO Range 200 to 6400

100 to 1600

Extended ISO Range 100-160, 10000-25600 50-80, 2000-6400
NEF Buffer (FX) 16-20 images 21-34 images
JPEG Buffer (FX Fine Large) 52 images 44 images
Minimum Advertised Price (US) US$4999 US$7999

If that table didn't tell you everything you need to know, well, I'm willing to keep writing...


Not Recommended (most)





This will be a controversial rating, to say the least. It really all boils down to value. If you need the extra pixels and have clients that'll pay for it, you assess the value differently and the D3x will make the cut. But for most users, the presence of the A900 and 5DII and the lower cost of the 1DsIII make the D3x not only the premium-priced player, but one that's currently too far out the margin in terms of value. Yes, I know the A900 and 5DII aren't necessarily the same body quality or even quite the same image quality, but the issue really boils down to this: you're paying for pixels (otherwise the D3 would be more expensive). Are the pixels "good enough" to justify the big price differential? No, not for most shooters. That doesn't mean that they aren't good, arguably even better than the competition. But the pixel differential isn't enough to justify the cost differential.

Nikon needs to put this sensor into a reasonably-priced D700x, and soon. Otherwise, the enthusiast market will likely consider migrating away from the F mount if it decides it wants more pixels. Ironically, that, too, will make the value proposition of the D3x much less attractive, just as the D700 made the D3 much less attractive to the larger segment of the market.




There are no Basics or Handling sections to this review because in terms of camera function, design, and handling, the D3x is the twin of the D3. Read my review of the D3 for comments on these things.

Build Quality
My original D3x came from Nikon with the viewfinder offset by about 40 pixels horizontally. If I frame by viewfinder, this represents a 1.5% mistake in framing (both edges are wrong, so I've got closer to 80 pixels that aren't as I want them). This is, by far, the worst 100% viewfinder alignment problem I've found to date on more than a dozen Nikon pro bodies. It is a real and unwelcome problem that just shouldn't happen on a US$7999 camera coming from the factory.

Several people reading the original review wrote to me and wondered if my viewfinder was out of alignment because of "rough shipping." No. In over a decade of using Nikon professional bodies and treating them with a lot of abuse in the field, I've never had a viewfinder mask shift, nor do I know any other photographers who have. I'm pretty sure that I subject my bodies to far more abuse in my travels than they get in shipping to me in their original packing. If rough shipping were to cause any significant viewfinder shifts, Nikon pros everywhere would be complaining about alignment problems. Thus, I'm of the opinion that a misaligned viewfinder found immediately upon unpacking a new DSLR would be a factory quality control problem. [After I wrote this comment in my original review I received comments from others who had received D3x bodies with viewfinder offsets, though none apparently as bad as mine.]

Large viewfinder offsets will impact autofocus, too, as the viewfinder points don't agree with the autofocus sensor positions. I'd judge the offset in my original D3x to be just below the point where it impacted autofocus significantly.

I've long been an advocate of everyone doing tests of their equipment to understand how it does and doesn't perform. Someone purchasing a D3x has an expectation of "high performance." In this case, I did not receive the kind of performance I expected, and I found that problem within taking 20 shots on the camera. Sure, I could have sent the camera back to Nikon for adjustment. But this gets back to something I wrote when the D3x was launched and will touch on later in this review: when you effectively raise your prices and become the premium player in the market, you need to make sure that you do what it takes to earn that. I made the suggestion shortly after the launch, for example, that a D3x ought to automatically register into the NPS Priority Repair program and/or Nikon needs to consider loaners when a product of this price and caliber is not up to expectations and needs repair or adjustment. That would have taken some of the sting out of the price. Consider if I wasn't a member of NPS what my US$7999 purchase would have produced: I'd either have to live with the pixel offset or I'd have to live with Nikon's regular repair cycles, which means I could be without my camera for more than a week, even for a simple adjustment. My US$7999 camera would also be subjected to a lot more handling and shipment than it should, and it would cost me more to send and insure that shipment back to Nikon. I'd be out time and money for something that should have been right on a US$7999 product.

People keep commenting about me writing that the D3x is overpriced. That is incorrect. I don't like the price, sure. I don't know anyone that does (other than perhaps Nikon). But my complaint isn't about price, per se. My complaint is about how you market and manage a high end product that's 60% more expensive than your previous high-end product. There's a problem that the entire tech industry tends to fall into: that better products automatically demand more price and you don't have to do anything else but make it better. That's just not true. As Toyota understood when they introduced the Lexus brand, it wasn't just about making a better, more expensive product, it was also about making the customer experience with the product better that commanded the dollar differential. My experience with my first D3x was worse than that with my D3, yet I paid 60% more for it. Oops. Worse still, it appears that my experience isn't an isolated case. I've received multiple emails about other D3x out-of-box experiences that were, well, less than satisfactory.

This is a subtle issue that eventually wears away at brand identity. The one thing you can't do if you increase prices is you can't make the customer experience worse. Now, my experience and those of the others who've emailed me may be random events, and for Nikon's sake, they had better be. If there's any sense that quality control in manufacturing has gone down while price has gone up, the US$7999 price would be a long term disaster for Nikon's reputation.

To Nikon's credit, they authorized an RMA for my original D3x through my dealer without knowing who was doing the return or why, and my dealer shipped me a replacement. This still has me out of pocket for the cost of shipping and insuring the package, though. For another photographer whose D3x sensor failed, Nikon overnighted him a new camera.

Okay, enough with the marketing and customer relationship lecture. Let's get to the good stuff: how well the camera performs.




top | home

No D3x Guide

I won't be doing a specific D3x Guide. At some point I'll update the D3 Guide to reflect the minor changes the D3x makes, which really boils down to different numbers in a few tables and some different comments about a few image quality aspects.




Another Do Over note: I've now run tests on four D3x bodies: my original one, my replacement, a friend's body, and my assistant's body (which has the AA replaced with a plain filter). In some tests, like those involving color, I found no real significant differences between the bodies. Ditto with resolution tests (other than my assistant's body without the anti-aliasing filter). On noise and dynamic range tests, I did find some differences between my original body and the other three. Since the three samples other than my original all fall in a very narrow range together, my descriptions here will use those bodies in the noise and dynamic range testing. And by reader demand, the basketball gym photos are back.


The D3x reminds me a lot of the D2x. At the base ISO there's a wonderful aspect to the image quality that's difficult to put into words, but is extremely impressive and not matched by any other current DSLR (I wrote something similar when the D2x came out). Acuity is high, dynamic range excellent, contrast is good and almost three-dimensional with excellent micro-contrast, and even though there can be noise in the deep shadows, this is well controlled and more film grain like than those digital color dots we're used to. I look forward to putting some real landscape challenges to this camera at the base ISO.

I'm slightly less impressed as I run up the ISO scale. That's not to say that the D3x is a noisy camera. It's that the noise eventually impedes upon the thing that you're probably looking at the camera to capture: resolution. As with my D2x, I'm going to have an ISO value I don't go above with this camera, I think. At the moment, that ISO value is 800. The issue is the deep shadows. In landscapes that tax the dynamic range, I'm finding that the very deepest shadows are indeed impinged upon as I bump ISO. At ISO 800 I've lost two stops of dynamic range, for example. Since the camera seems perfectly fine for long exposures measured into the low minutes, I'll take a longer exposure over higher ISO and keep the little bit extra in the shadows, just as I did with the D2x.

Some people reading my original review saw the reference to the D2x as an indictment. Indeed, I even alluded to that myself. With a little more distance from my initial impressions, I wouldn't put it that way, though. What I meant by the references to the D2x were mostly positive thoughts. To this day, I think I'd slightly prefer the D2x at base ISO to any of the other Nikon 12mp cameras. The problem with the D2x is that image quality degrades very quickly as you bump the ISO values. That's not really true of the D3x. From a pure noise standpoint, it holds its own with most any current DSLR. Indeed, at 14-bits the D3x is only slightly behind the D3/D700 at ISO 3200. The same can't be said for the D2x. Thus, I've toned down my D2x reference a bit so that it is more clear what I was writing about: at base ISO the images out of the D3x are superb, and they remind me of the joy I had in seeing D2x base ISO images for the first time.

As good as the D3x is--and it is, again, in my opinion it is the best DSLR currently made--the D3x is not quite the be-all, end-all camera some think it is, though it comes closer than many might think. Moreover, extracting that best DSLR performance is going to tax the skills of a lot of users. If you haven't performance tested your lenses and support system, the D3x will force you to do so.

One final comment before moving on. While I'm going to try to show a few of the things I write about, remember one thing: in order to keep bandwidth usage down to a reasonable level, I never put full-size, high-resolution images on this site. Perhaps when I get around to redesigning and moving the site to another server I'll readdress that, but for the time being I have some caps I need to try to live under. That means that what you see here may not be exactly what I'm seeing (mostly due to higher compression and the use of JPEGs). As usual, go by my words, not necessarily the samples as you see them. I've tried to use samples that will show some of what I'm writing about, but often these still mask some of the very fine differences that are in the native raw files of my tests.

I've harped about Nikon's marketing of the D3x elsewhere, mostly about the way it was unveiled, but I do think it's worth another close look at some of the Nikon marketing claims before we get to the actual results:

"...breaking new ground in imaging quality." A big claim. Nikon essentially is saying that the D3x has image quality better than any other DSLR. The bar was already set pretty high by the Canon 1DsIII a year ago, and the Canon 5DII and Sony A900 have echoed that level more recently. You can probably already guess my answer here, but the short version is that this claim is justified.

"...will raise the bar for commercial, fashion, and stock photography." Here's where things get somewhat muddy. Many commercial and fashion photographers shoot medium format. Is Nikon saying that the D3x raises that bar? That the D3x is better than MF? I think you'll find this marketing claim is a bit of an overreach. The claim should have included a "for DSLR users" clause and then it would have been accurate.

"...without a doubt our highest quality camera to date." This claim is well justified. Certainly at the lower ISO values there's no disputing it. Only at the highest ISO values would there be any contention, and even then the D3x does well enough to clear the bar. [Note: in the original review, I had the word "the" instead of "our". This obviously changes the context, so I've changed my comment.]

I want you to keep those marketing claims into mind as we enter into the more detailed look at the image quality the D3x does produce.

As I've come to expect from Nikon cameras, the color rendering is generally excellent, if a bit over saturated at virtually all of Nikon's Picture Control settings. Here's a typical analysis for the Neutral Picture Control at ISO 100 (left) and ISO 1600 (right):

Copyright 2009 Thom HoganCopyright 2009 Thom Hogan

The midrange and reddish colors tend to be about as close to dead on as you'll find in a camera, the blues and greens exaggerate a bit. All of the Picture Controls other than Neutral tend to push colors a bit more towards the boundaries. Automatic white balance does a good job within the narrow range I have available in my studio (basically 4000-6000K), being slightly under the actual Kelvin value but in a fairly narrow range of miss that I find highly acceptable. Surprisingly, boosting the ISO doesn't have quite as bad an impact on color as I would have expected with the small photosites of the D3x: the saturation is still high, and the color errors haven't increased as much as we've seen with some previous Nikon bodies. The D3x, like the D3 and D700, tends to "hold color" very well into the higher ISO range, an impressive performance.

I am, however, noticing a tendency for the D3x to blow out red. I was surprised when three colors on my ColorChecker SG chart consistently would blow the red channel at the matrix meter's indicated exposure (but not with my D3 in the same test setup). Further investigation shows that the D3x has very strong red performance: this is not the same Bayer filtration that sits on top of the D3 or D300 sensor. This isn't exactly a fault, just a warning. Let the white balance drift too high in incandescent lighting and you could get tough to deal with Red channels if you're not careful. Landscape shooters need to worry about this at sunrise and sunset, too. But if you're capable of shooting with a camera of this caliber, I'm not telling you anything you wouldn't already be checking. I guess my real warning is this: the channel balances on the D3x may be a little different than you're used to. If you think you've got channel balance nailed for your current exposures on, say, a D2x, you may find that you need to tweak your exposures or white balance a bit on the D3x. Of course, if you're using UniWB, you're just looking at the channel histograms all the time anyway, so it isn't a big deal.

Overall, there's nothing to fault in the color rendering of the D3x. It is accurate, predictable, and stays in a very tight range even when you boost ISO values. Moreover, color rendering was consistent amongst the D3x bodies I tested.

Okay, things are going to change quite a bit from the initial review. Consider just these numbers:

  Red Green Blue Luminance
Original D3x ISO 200 .89% .86% .88% .82%
New D3x ISO 200 .79% .72% .77% .71%

These are the same Imatest data, using the same lighting, using the same camera settings. Unfortunately, because I had to return my original body to get the new one, I couldn't test them side by side, but my test rig doesn't change, and I generate multiple images at the same settings each time I test to rule out lighting shifts and other issues. Temperature was 5 degrees F higher in testing with the new D3x, which should put it at a slight disadvantage. The exposure difference according to Imatest was within .03 stops between the two. I'm pretty confident that the results listed above are indeed different due to the cameras, not my testing. To put the results into perspective: at ISO 200 there was a 13.4% improvement in luminance noise handling. At the higher ISO values, the difference was lower, but still significantly significant. (Why am I testing at ISO 200? Because I was doing direct comparisons against the D3.)

When Lloyd Chambers first started reporting some of his results from his D3x testing, I was scratching my head. I didn't see the same thing with my camera. Well, now I know why I wasn't seeing the same thing: sample variation. Frankly, I'm surprised to see any sample variation in a US$7999 camera assembled at Nikon's best plant in Japan and obviously targeted at professional users. But having now tested four bodies and finding one to be "off" in several ways, I think I have to go with what the tests say. For those who are into the details, the serial numbers of the two bodies in the above chart were about 2000 units different and the bodies were shipped to my dealer from NikonUSA about a month apart.

Even the original D3x had a strong noise performance, and a surprisingly balanced one. But the replacement body is just better. It was apparent to me the first time I pulled up an image on my display that something was a bit different. And the testing tells me what it is: tighter noise handling. At the base ISO the (new) D3x produces numbers as good as I've ever seen. As we ratchet up the ISO values, there's definitely noise, but very well handled noise.

You're probably wondering what all this means in terms of visual impact. Okay, here we go down the rabbit hole at 100% view:

Copyright 2009 Thom Hogan

Copyright 2009 Thom Hogan

The top image is ISO 100, the bottom ISO 3200. One of the things that's definitely in play is brightness. The darker the patch, the more noise becomes obvious. This is generally true of all digital cameras, but it seems the ramp up is a little more for the the D3x than I see in, say, the D3. Underexposure is absolutely not your friend on a D3x, and deep shadow detail is going to be hard to distinguish from noise at higher ISO values. Note, too, that the perceived colors are shifting darker: the non-noise is still retaining the right color, but the noise all tends to be just a darker relative of that color. The good news is that there isn't a lot of that nasty color noise we often see at high ISOs (though look at the gray patch at right: you should see that there is definitely color noise starting to come in here at ISO 3200). Instead, most noise out through ISO 1600 is very much luminance noise, and only beyond that am I seeing a stronger ramp up of color noise, too.

One of the discussions that I kept seeing when the D3x first got into users hands was how it was "better" than a D3 at ISO 3200, that the extra pixels somehow made up for noise values. It's better than I thought it would be, but still not quite what people were claiming. Let's look at some detail at the two ISOs to see why (ISO 100 on left, ISO 3200 on right):

Copyright 2009 Thom Hogan Copyright 2009 Thom Hogan

I'm using a section of a standard resolution chart that's at and beyond the Nyquist frequency. At ISO 100 (left), the structure of what the D3x can and can't resolve is easily seen. At ISO 3200 (right), the structure gets quickly masked by noise, and there is very low acuity on the edges. An additional problem is this: to take the noise out, your NR routine will take some more of that resolution away (all the previous samples are with NR off, by the way).

Here's the same ISO 3200 sample (left) with noise reduction applied (right):

Copyright 2009 Thom Hogan Copyright 2009 Thom Hogan

Personally, I think you either live with the noise the D3x produces (which again, is mostly luminance and not particularly objectionable by itself) or you use a D3 at the highest ISO values.

Since so many people asked for it, I've reshot the D3 basketball gym shots so that I could compare against the D3x. This was necessary because they changed the lighting in the gym substantially (at least a stop more light, a bit different in color, and more frequency based than before). So first things first: you can't compare the following shots with any that appear in my previous reviews. Only the comparison here is apples to apples. ISO value is now 1600 instead of 3200. You'll note that the focus is slightly different in these two shots (I have to shoot them during short breaks during the game, and sometimes my sweat makes it a bit difficult to see through the viewfinder well): the D3 shot is focused on the front of the rim, the D3x is focused on the back of the rim. So if you were trying to evaluate detail, you need to take that into mind.

First, here's a 100% view of the cropped D3 shot, which has been resized to match the size you'd see with the D3x (I haven't done anything fancy with the resizing, just a quick Bicubic in Photoshop):

Copyright 2008 Thom Hogan

Now here's the D3x shot (no resizing, 100% view):

Copyright 2008 Thom Hogan

These shots actually do a pretty good job of illustrating the difference between the two cameras at high ISO as it turns out. The D3x gets more detail (note the texture in the in-focus portion of the net strings), but it comes at the expense of noise in broad same-toned areas. The question then becomes how much do you lose by doing noise reduction (I've used Neat Image tuned to not be very aggressive to produce the next image):

Copyright 2008 Thom Hogan

The answer is that we've lost detail in the strings. I wouldn't judge the D3x image to be better or worse than the D3 image at this point. There are some very small things that are better in the D3 blowup image, some things that are better in the D3x noise-reduced image. So at ISO 1600 I'll call it a wash in my gym test bed.

So the question some ask at this point is this: why shouldn't I shoot with the D3x if I can get results just as good at high ISO? Well, first we have the issue of workflow. The camera's built-in noise reduction is decent, but not optimal. If you need to add a post processing noise reduction step, then you might increase your workflow burden by using the D3x (likewise, blowing up a D3 image might increase your workflow burden, too). A second concern is dynamic range. At ISO 1600 as in this example, the difference isn't particularly meaningful, but because we tend to lose about a stop dynamic range with each stop boost of ISO and the D3x lags the D3 from ISO 400 onward in dynamic range, at some point you're going to be contending with a "block up" of values and want whatever small gains you can get. I personally judge that point to be somewhat above ISO 1600 for me. Below that, I'll take either camera; above that, I want my D3.

Still, one has to be impressed with the D3x's abilities at high ISO values, especially from in-camera JPEGs, which seem very well tuned (put another way, it's difficult to get better high ISO work from NEFs than the camera's internal conversion does).

Okay, we've got the good noise news out of the way, now for some bad: my original D3x had excessive amp noise, my new D3x has excessive hot pixels To illustrate just how bad it can be, here's a 15-minute exposure black frame exposure at ISO 800 with noise reduction off (I've adjusted levels very slightly to make it more obvious in these reduced size samples):

Copyright 2009 Thom Hogan Copyright 2009 Thom Hogan

On the left, we have the original D3x I was sold. On the right, we have the replacement. In all the long exposure testing I did (different ISO values, different exposure lengths, different Long Exp NR settings), the original D3x was worse than the new one, and I could detect that visually without even having to measure the images.

However, all is not perfect with new model. Here's a 100% sample of the lower right corner:

Copyright 2009 Thom Hogan

(ISO 800, 15 minute exposure, Long Exp NR Off, Ambient temperature 76 degrees F)

Long exposure noise reduction helps remove the hot pixels, but there's still a small amount of amp noise creeping in on almost all the image edges of my new D3x at ISO 800 and 15 minutes. Fortunately, at the exposure ranges I'm likely to use (ISO 100, up to five minutes), the results are very acceptable and on par with what I would expect. I can live with the D3x's performance here, but I was hoping for a little better.

Equipped with new test charts, I set out to illustrate the resolution differences between the D3 and D3x. Instead of writing much about numbers here, I'm going to try to give you a real world sample. Here's the overall scene taken with the D3:

Copyright 2009 Thom Hogan

The lens is the 300mm f/2.8 mounted on my heavy tripod. Focus is obtained via Live View on the resolution chart. I like my neighbor's garage (and will pay him not to paint it) because the texture in the weathered paint on the door is exactly that kind of random but repetitive detail I encounter all the time in landscape shooting.

Okay, back down the rabbit hole. Let's see how you do with this Rorschach test:

Copyright 2009 Thom Hogan Copyright 2009 Thom Hogan

Copyright 2009 Thom Hogan Copyright 2009 Thom Hogan

Before I reveal what those four images are, do you have any preferences? I'll give you a hint: two are from a D3, two are from a D3x. Can you figure out which is which?

Probably not, because there's another thing going on here. The top row are D3 images, the bottom row are D3x images. The left column are images with the AA filter intact, the right column are images with the AA filter replaced with plain IR-blocking glass.

Here's another piece from the same image:

Copyright 2009 Thom Hogan Copyright 2009 Thom Hogan Copyright 2009 Thom HoganCopyright 2009 Thom Hogan

On the left are the D3 pairs, on the right the D3x pairs. On the D3 bodies, the Nyquist frequency is somewhere around the 4, while on the D3x it occurred somewhere between the 5 and 6. Note that the D3 with its normal AA filter (leftmost example) produces very little color fringing. Note that the cameras without AA filter (the second and forth example from left) produce a wider range of color fringing.

Okay, let's look at things a little differently:

Copyright 2009 Thom Hogan Copyright 2009 Thom Hogan

What do you think we're looking at now? I'll give you another hint: one of the images is a D3x image, the other is an upsampled D3 image. Go ahead, give it a moment and decide which is which before reading on.

If you said the left one was the D3 image, you'd be wrong. What happened? Well, this is a point I've been trying to get across to people: just using a camera with more pixels isn't enough to get more resolution and acuity. In this case, the D3x was slightly off focus (and I mean slightly). I also was extremely careful in how I upsampled and sharpened the D3 sample on the right.

Put another way, it takes a great deal of care and attention to detail to extract everything the D3x is capable of providing.

Since I know you'll ask for it, here's another pair (better, or worse?):

Copyright 2009 Thom Hogan Copyright 2009 Thom Hogan

Hard to tell which is which, isn't it? (D3 upsampled on left, D3x on right). One thing you need to consider in the D3/D3x choice is whether the subject matter you're shooting actually has detail that's going to reveal the resolving differences of the cameras. And, of course, diffraction and lens choice also come into play, as well.

There's no doubt that the D3x is the resolution champion. It clearly outscores anything I've tested before (and I went back and did some retesting with the new charts just to be sure). But the range of mistakes you can make to get that resolution is considerably narrowed. I found a small amount of tripod mount play on a couple of my lenses that showed up clearly on the D3x but not when using my D3, for instance. Out at f/11 and above the diffraction starts to clearly take back that resolution (see diffraction, below). Some lenses, used wide open, don't quite have the resolving power that the D3x does. Small focus mistakes take away resolution more clearly (time to learn how to use AF Fine Tune). And so on.

Finally, the D3 seems to me to hold its resolution well at ISO 1600 and beyond, the D3x, not so much. One of the things that is confusing some people is that there appears to be more detail in the high ISO D3x images, but there isn't. The dithering impact of the D3x noise fools you into thinking that you're seeing very small detail. But you're not. You actually are seeing faux detail in the noise patterns, much like removing the AA filter on your camera often gives you faux detail above the Nyquist frequency. Some people like that visually--I happen to agree, as long as the noise has no color component to it--some people don't. But the reports of the D3x out resolving the D3 at very high ISO values are, I believe, wrong. I measure very substantial resolution drops on the D3x on test charts at high ISOs that I don't get from the D3. I judge the crossover where the D3 measures better at most JPEG settings to be somewhere around ISO 3200, at least for resolution.

My bottom line is this: through ISO 1600 I'll take the D3x. From ISO 3200 up I'll take the D3, no questions asked. In the range between those two numbers, I'd tend to lean towards the D3 at ISO 1600 but wouldn't automatically reject the D3x. There's more going on than just what I showed here: some of it has to do with the fact that I shoot raw and use some specific workflows to "optimize" my pixels. In camera JPEGs using Nikon's in-camera noise reduction might let me shoot a little further into the high ISOs with the D3x without getting upset by what I'm losing.

(It was pointed out to me by another tester I trust that there may be light components to noise and resolution differences, especially at high ISO values. For instance, in incandescent light, the D3 starts to look better at a lower high ISO value than under daylight. That's starting to really pick nits, but it is something you should consider if you're shooting indoors all the time under artificial light: the D3x may not quite give you the same performance at high ISO values as I get with my mostly daylight shooting.)

Dynamic Range
My comments here are new and revised. At base ISO and 14 bits I'm finding that the D3x has perhaps a full stop advantage over the other ISO 100 Nikons. Best case measurement on the D3x just tops 10 stops, the best I've measured on a Nikon DSLR. Even at ISO 200 the D3x still seems to have a third to maybe a half step advantage over the D3 and D700. At ISO 400, the D3 and D3x are about equivalent in dynamic range. By the time we get to ISO 800, the D3 has a slight advantage that continues out to the maximum ISO 6400 of the D3x. The interesting thing is that the D3 and D3x are clearly in a different class than the DX sensor bodies in terms of dynamic range. A D3x at ISO 3200 has about the same dynamic range as the D300 does at ISO 1600. That's pretty remarkable, as the photosite size of those two cameras are relatively close. I was a little surprised by these results. For my work, the extra dynamic range is welcome. But this also points out how the D3x does so well in high ISO noise compared to the D3. Dynamic range is, after all, the measurement between well saturation (photosite blowout) and noise floor. Nikon has done something pretty remarkable here.

Diffraction is a contentious and complex subject. Contentious partly because of misinterpretation, and I'll be the first one to admit that I've used the term "diffraction limited" a bit too casually in the past. Complex because multiple things contribute to whether you'll see the diffraction impacts in your image.

As you may recall, diffraction is produced by the opening of the lens (aperture opening). As such, the actual diffraction is always "the same" at f/16, no matter what camera you put that lens on. The matter gets a little complex due to whether or not the camera is recording the diffraction accurately. It should be obvious that if the diffracted light still falls onto the same photosite as the non-diffracted light, you wouldn't actually record the diffraction. What's not so obvious is what happens with multi-layer anti-aliasing and Bayer filters sitting on top of that photosites, and demosaicing that "interprets" actual pixel values.

In the past, I've used the term "diffraction limited aperture" to mean the aperture at which the diffraction is essentially fully recorded. Any physically smaller aperture than that would have the predicted additional diffraction impacts on the visual image. So if I wrote that "the D2x is diffraction limited at f/11" that would mean that f/11, f/16, and f/22 would be fully recording diffraction impacts, while f/8 would only be partially recording the expected diffraction impact. That doesn't mean that f/8 didn't have any visible diffraction differences in it from f/5.6, it just meant that the mechanics of the system weren't fully recording the diffraction yet.

So this time I'm going to give you the full range of resolution impacts using a known good lens. It's important that you see where and how much change there is as you stop down, because a lot of folk on the net are making comments about the "resolution" ability of the D3x without taking everything into account. If you shoot a low-light test at f/2.8 you're going to find something different than if you shoot it at f/11 in bright light.

In the following examples I've tried to produce "best case" diffraction examples. These are center area 100% samples from a 200mm f/2 lens mounted onto my best support system and triggered via remote with mirror up. I've set in-camera sharpening to a modest mount (5), and I've increased contrast by resetting and clipping the white and black points. With care in post processing sharpening of a NEF, I can get a tiny bit more definition out of the f/11 sample, but I think this set of images is pretty indicative of best possible results for most situations. I should note that the converging lines on the right of these samples go right to about the Nyquist frequency expectation of resolution for the D3x. My expectation was that only diffraction-free apertures would resolve the smallest lines here.

Do Over note: I've gotten new charts and spent a lot of time trying to get everything locked down tighter than the previous samples. One thing I discovered in this process was that part of the problem I was having with the slower shutter speeds at f/22 were due to the tripod collar on the 200mm lens. At the rotation point for this collar there's just a tiny bit of play, so what happens when you get any vibrations in the support system is that this play starts to show up as camera shake.

I've also simplified the examples here to just show a single best case aperture against the apertures where diffraction starts to come into play.


Copyright 2009 Thom Hogan This lens is near optimal at f/4, but f/2.8 comes very, very close, as does f/5.6. We'll use this sample as representative of "best we can get." Note that the blacks are intense and there is very little gray and/or color between the smallest converging lines.
Copyright 2009 Thom Hogan At f/8 contrast has started to reduce (whites are slightly darker, blacks are a bit lighter). Between the smallest converging lines there's a bit more "fuzz" happening, but this is still quite good performance.
Copyright 2009 Thom Hogan f/11 is probably the smallest aperture where I feel I can still pull up some very fine detail. We've lost some more contrast, though, and edges now have a slight fuzz to them that clearly indicates diffraction is present.
Copyright 2009 Thom Hogan Things have taken a big turn for the worse. Fine detail is now masked and edges are distinctly fuzzy. Sharpening is actually beginning to be counterproductive.
Copyright 2009 Thom Hogan At the smallest aperture on this lens we've lost all edge integrity.

To be consistent with what I've written about other Nikon bodies, I would say that f/11 is that last aperture you can use and still have an expectancy of getting near maximal resolution (a term I loosely used before: diffraction limited aperture). Go smaller than f/11 and no amount of careful sharpening will restore the semblance of acuity at and near the Nyquist limit. Moreover, you'll lose significant contrast, and most people perceive contrast in conjunction with sharpness (i.e. more contrast is perceived as "sharper"). If you're a stickler for detail, don't stray past f/8. While I've only demonstrated one set of tests here with one known good lens (and sample checked against another), what I just wrote appears to apply to all the high quality lenses I tried on the D3x. For my landscape work, I won't be afraid to use f/11, but if I can get to f/8 I will.

What if your lenses aren't as good as the 200mm f/2, 24mm PC-E, 50mm f1.4G, and other excellent lenses I tested? Well, the D3x is going to reveal your lenses' flaws, then. There's been a lot of anecdotal and casual "tests" that have been posted on the Internet, some of which conclude things like "I don't see any big difference between f/9 and f/16 on my Bizarro 31-89mm f/3.7-6.3 shot handheld." Well, neither do I ;~). One reason I shoot test targets and calculate out known values (Nyquist, etc.) is so that I know what it is I'm looking at. I'm not perfect at that--even the best tester can get slightly different results in retests--which is why I shoot multiple tests with multiple setups and multiple lenses. When I'm comfortable that what I see is consistent, I'm confident that my results are projective. Thus, I have strong confidence that f/8 is as far as you should go if you're an absolutely stickler for best possible results (and it may come earlier than that with some lenses whose performance peaks at a lower aperture), while f/11 is a good compromise point at which the sacrifice is small and potentially well worth the depth of field gain.

This is not to say that I never shoot at f/16 or smaller on a camera like the D3x. But it means that I know I'm compromising absolute resolution for something else. That "something else" needs to be something pretty tangible, and I need to know that I need it. Put another way, I won't casually wander to an aperture smaller than f/11 on the D3x. Ever.

Frame Rates and Buffer
I can't complete a performance section without writing about continuous shooting. There's good and bad news here. If you're a 12-bit NEF shooter, you'll be happy: the D3x actually buffers more shots than the D3/D700 do and manages to give you a respectable 5 fps while doing it. JPEG shooters aren't likely to encounter buffer issues, either. It's only if you shoot 14-bit NEFs that things turn downwards. The minute you dial in 14-bits, you lose frame rate. For the FX frame that means 1.8 fps.

I don't do enough studio shooting with models to be able to speak authoritatively about whether that's "good enough" or not. It feels like not to me. I have shot with some fast medium format (MF) choices just enough to say this: I don't think the minimal frame rate increase of a D3x over many MF bodies makes up for the loss of resolution on current state of the art MF. At 5 fps, sure, if you've got the lighting to handle that frame rate, the D3x starts to give you flexibility that you don't get with the MF options and thus feels like it offers something very valuable: spontaneous and near continuous shooting. But at 1.8 fps, I question how much you're getting over MF. Yes, the focus system might be better on a D3x than any available MF camera, but it doesn't feel like that's enough of a benefit. I'd rather have more pixels and the other advantages of MF, I think.

As I said, I don't do enough studio shooting to fully trust my reaction there. Perhaps I'm missing something that's important to a studio shooter. So if you fall into this category, here's what I'd say: rent a D3x and do your own comparison before buying. You're trading off resolution and price against a lot of other factors (including leaf shutters in some systems that allow you to break out of the restricted shutter speed ranges of DSLRs with flash). I can't make the final assessment for you, as each of you will see value in different aspects (arguably, the D3x has a better feel in hand for continuous shooting than the Hasselblad H3 I used briefly, for example).

For the landscape shooter like myself, I don't see the frame rate drop to be a particular issue (though it would increase HDR capture times, which can be a problem with wind and any subject motion it causes). And yes, I see just enough difference in the really deep shadows to warrant shooting 14-bit for my usual subject matter. The way shadow detail is recorded is a very subtle change, though, so if you need the frame rate, I'd say don't be afraid to pull back to 12 bits. However, we're still early in D3x-aware converters, so I'm not yet sure of how much more I can extract from 14 bits over 12 bits.

(So what's happening with 14-bits versus 12-bits? I don't think the electron well of the D3x photosites is big enough to fully justify anything larger than recording as 12 bits, so it isn't that you're missing out on data. I believe that's what's really happening is that the method Nikon is using to trigger the 14-bit data stream is doing something that makes that the lowest order data slightly more accurate [or reliable, take you pick of the wording]. As I said, the results are very subtle, but 14-bit shadow noise seems slightly easier to work with than 12-bit shadow noise.)

Lenses and the D3x
I've seen a lot of chatter on the Internet about how the high resolution of the D3x means you need only the highest quality lenses. I'm not sure that's perfectly correct. I find it interesting that each Nikon camera lately has had a slightly different "lens personality." This is coming from the different AA filter choices and changes in microlenses, I think.

Many people forget that there are optical changes happening right at the sensor. Your lens focuses light to a precise point, but from the front of the AA filter to the sensel collecting the light there's actually a measurable and significant distance on DSLRs, and both that distance and what's happening as light traverses that space is different from model to model. So, how a lens presents light to this sensor area optical system is important.

Most telephoto lenses I've tried on the D3x look as I'd expect. Low quality ones produce less than ideal results, high quality ones make for impressive resolution results. I was once again reminded of just how good the 200mm f/2 and 400mm f/2.8 are, for example. To a large degree, the same is true at the wide angle end, as high-end recent designs with the post 17-35mm style rear element design do quite well, while earlier inexpensive designs show clear edge to edge issues. The 24mm PC-E, 14-24mm Nikkor, and my Zeiss 18mm and 25mm lenses all do quite well, while my older 20mm f/2.8, 24mm f/2.8, and several other lenses look much weaker as you move from the center of the frame outward (especially in raw files, where chromatic aberration reduction isn't done). This is all to be expected. It's in the middle focal range where I've been a bit surprised, with some lower cost lenses outperforming higher end lenses for no reason that I can really explain (and I don't think it's sample variation). I'll need to continue testing to see if I can figure out what's going on. But one thing you need to be cognizant of is whether your lens is flat field corrected (CRC) or not. Some of the issues I'm seeing reported for lenses on the D3x aren't really "soft corners," but rather lack of flat field.

That said, yes, in general the D3x does highly reveal the faults in a lens if it has any. Even the vaunted 14-24mm and 24-70mm on the D3x reveal that they're not perfect into the corners as some have thought using D3 and D700 bodies. If you're a pixel peeper and trying to extract the largest possible print out of your new D3x, be prepared to do a lot of lens re-evaluation. If you want to take advantage of all those pixels, you need to make sure that both what you put in front of them and how you handle the camera are up to the job. At some later date when I've had a lot more experience at this, I'll probably add some more comments about individual lenses.

I should also point out that you really need to consider using AF Fine Tune (or contrast based Live View focusing). Just playing with my D3 and D3x and dialing in correct and "off by 10" values revealed that I'd rather have a D3 tuned properly than a D3x that's off. At least for test charts ;~). But that brings up another issue. On my original D3x I wasn't able to "dial in" my 200-400mm at 400mm, though I was able to do so on all my other bodies that have AF Fine Tune. Unfortunately, on that D3x the "best 400mm" setting made the 200mm focal length unusable. I hope that Nikon thinks through the AF Fine Tune system some more. We need the ability to track individual lenses (it's not unusual for me and my assistant to swap our bodies onto each other's lenses sometimes, as that's easier to do than each of us completely doing a new setup while trading positions). We need the ability to tune for both near and far distances, for focal lengths on zooms, and for using TCs.

(Actually, this is posing a bigger problem for me that I'm going to have to think about. In some cases, the same lens on a D300, D3/D700, and D3x needs different commentary and perhaps even different ratings. Right now my reviews don't reflect that, and I'm not yet sure how I should proceed to add that this site.)

Image Quality
All of the above makes for a rather complex set of inputs necessary to discuss image quality. Let me illustrate:

  • Fixed focal length telephoto lenses at maximum aperture: Superb. Most of the exotic Nikkor lenses (200mm f/2 to the 600mm f/4) are designed to produce best results wide open or close to it, and they do on the D3x. We're not usually stopping down much if any using these lenses so diffraction doesn't rob us of any contrast or resolution. Generally we aren't too worried about the extreme corners for subjects we shoot with these lenses, but these lenses tend to be darned good into the corners, too.
  • Fixed focal length wide angle lenses at minimum aperture: Very good. When we start doing landscapes at f/11 we're starting into diffraction territory. Indeed, I think that the serious landscape shooter almost certainly needs to consider using the PC-E lenses on his D3x and using the Scheimpflug principle to get depth of field, not solely aperture. If you need to go to f/16 for depth of field in your landscape work, I'm not so sure that you shouldn't just use a D700 and use panorama stitching to get more pixels, despite what that means for moving subjects.
  • Fixed focal length mid-range lenses at maximum to moderate aperture: Excellent to Superb. You'll find that most of those lenses aren't quite as good wide open as you thought, as the D3x will show that stopping them down a stop or two gives you more contrast and resolution, especially in the corners. The D3x will do surprisingly well at higher ISO values, too, as long as the light isn't too skewed away from 5000K.
  • Wide angle and mid-range zooms: Good to Excellent. The two most recent offerings, the 14-24mm and 24-70mm are at the excellent end of this range, some older ones like the 17-35mm tend much more towards the good side.

Starting to see the problem? There isn't an easy, single "image quality" statement I can make. How you use the camera will determine how good you think the image quality is. What I can say is, avoid these things:

  • Small apertures. Once you get past f/8 you'll be fully recording diffraction, which starts to rob the D3x of one its seminal properties: high resolution. Not that you can't shoot at f/11 or even f/16, but you don't quite get the full benefit of the extra pixels that you'll get at, say, f/5.6.
  • ISOs above 1600. Again, the D3x is surprisingly good at high ISO work, but there's a point at which I feel that the returns aren't worth it and I'll use my D3 instead. Up to ISO 1600, I'll probably take the D3x. But one thing that starts to happen at ISO 3200 on the D3x, the loss of color differentiation and saturation, clearly starts to make those high ISOs less valuable to me. I personally will switch the D3/D700 at ISO 3200.
  • Low cost lenses you haven't tested. I've found a few lower cost lenses that do quite well on the D3x especially in the central region, but if you're looking for edge-to-edge performance with all those pixels, you want relatively recent, high-end lenses, especially as you move towards the wider end of the focal range.


The thing that stops most people cold is that the table I started this review with is there aren't a lot of key differences between the D3 and D3x cameras, but as I write this, there's a US$3600 price differential between the two at street prices. Thus, here in the US you pay about US$3600 extra for the things in the right hand column (and note that some of them are worse than a D3). To put it more bluntly, you're basically paying all that extra cash for double the pixel count.

So, before getting to the camera, we need to talk about the price. Nikon appears to have priced the D3x at the 1DsIII price (given that there are only two competitors, such a pricing decision can sometimes be found as an informal form of collusion in some jurisdictions). Essentially, Nikon appears to be saying that "if Canon's flagship is worth X then ours should be priced at X." The problem with that is that the Canon 1DsIII doesn't sell for US$7999, and pretty much has never sold at that price. It's fallen to a price basically just above what dealers pay for it, which is why many dealers don't stock the 1DsIII but will order it for you. In case that wasn't clear: the market demand for a high-quality US$7999 camera is very low, which is why the 1DsIII price has dropped so much and it doesn't sell in very high numbers. The market itself, therefore, has valued a pro body flagship with 20+ megapixels at a lower price than Canon intended it to be.

Meanwhile, the Nikon D3 price has also dropped down to near dealer price, indicating demand for it, too, is relatively low now. If D3 sales were soft, then what did Nikon think would happen introducing a camera with a list price US$3000 more?

Still, it's not the price, per se, that's the issue here.

I know I've been criticized by others for criticizing Nikon's launch of the D3x, but I'm not going to back down on this: it was just plain wrong strategy to just dump the D3x out with a press release and a US$7999 price. The higher the price, the more you have to do to set customer expectations in your marketing. Customers have long memories. The way the D3x introduction was handled is going to be remembered, and subsequent announcements are likely to analyzed even more carefully. If Nikon announces another camera with what appears to be a price premium, I think things will be settled in customer's minds: Nikon charges more. That raises the bar in telling the customer why you charge more. And as I've written several times over the last two months, the real problem with the D3x launch wasn't the price, it was that Nikon did nothing to set expectations nor did they make much of an attempt to explain or support their decision. The higher the price you charge for something, the more you need to handhold the customer and establish a close relationship with them. Nikon did neither. This is a classic marketing mistake.

Some have defended the price by saying "Nikon's costs on the D3x are more." Nonsense. The D3x has the additional buffer memory that is optional on the D3 (about US$500 cost to have installed). But the D3x doesn't have the very expensive ADC components that the D3 does, so the non-sensor parts costs are probably pretty close to a wash. So are we really to believe that the extra cost to Nikon of the new sensor compared to the D3 one is somewhere in the US$1000 range? (The US$1000 comes from the fact that a retail price differential of US$3000 is usually generated by a US$1000 difference in actual parts costs.) That seems not only unlikely, but well beyond unlikely. Nor does it cost Nikon US$1000 a sensor to put a gapless microlens array or a different AA filter on it.

I'll state with confidence that none of the actual parts costs justify the differential between the D3 and D3x. Thus, we're left to evaluate the camera based solely on image quality and customer support as to its higher price. As you may have noted in the first sentence of this review, and stated several times more, I think the D3x is the best DSLR currently made, and that's mostly based on image quality. But as that first sentence also intimated, just being "the best" may not be enough.

So, Nikon did one of two things (or both): they decided to try to sneak in a price increase, or they decided that their high-end camera had to have the same price as Canon's. As I've long written, Nikon isn't particularly customer friendly, nor are they particularly communicative. So we got what we got from them, take it or leave it. I've received numerous emails from people that say they'll leave it, and few that say that they'll take it. Perhaps this review will change the proportion, but there's still going to be a large number that will just skip the D3x.

Usually, I don't write much about price. Indeed, through much of the digital era Nikon has tended to be aggressive in price, which has made it not much of a factor to discuss. Unfortunately, the D3x has to be discussed relative to price. Had it been priced closer to the old US$4999 price point for high-end Nikon DSLRs, the whole discussion might have flipped: we'd be debating about whether it was a great value rather than a poor one. The terrible global economy doesn't help matters. At a time when everyone is looking to cut back, or at least find bargains, here we have something quite different: a very expensive tool at the top of the food chain and the price charts. So, regardless of what you think of the price versus the performance of the D3x, I think it has to be discussed. You may not agree with my assessment of value, but make sure that you understand how you came about yours.

I want to further clarify one thing. It seems that some people read what I've written and conclude that I'm disappointed with the D3x. Not really. For my work, the D3x does move things forward. I can now capture what I've been doing in stitches with one shot, and if I continue to use the same stitch techniques, I'm competitive with some medium format work. I'm perfectly happy to purchase a D3x, as it does appear to push my shooting into a realm it hasn't reached before. But my shooting is generally in a narrow range (low ISO, wide angle, careful setups, lots of post processing). While I'll be happy with the results I get out of my D3x (though not so happy carrying such a large, heavy camera into the backcountry), the point of this review is to try to cover more bases than just my landscape-centric viewpoint.

The question at hand is whether the D3x is a great value, a fair value, or a poor value. There will be a small handful that may say that the D3x is a great value. I'm not one of them. For my work, its somewhere in the fair range. Indeed, I think that's true for most professionals, too: the D3x is a fair but not great value. That's why I've assigned it three stars in ratings. (If I awarded half stars, I would have been tempted to give it three-and-a-half, though I'm not completely sure I would have.) Had Nikon used the pricing of the D3x to start a closer relationship with its highest end customers, I'm sure I would have made that rating higher. But instead, D3x shooters still mostly have that distant relationship with Nikon we've always had. Pity.

For the serious amateur and even professional that doesn't need pixels (photo journalists, for example), the D3x value isn't so good, which is why I rated it at two stars. For many, the Canon 5DII actually looks like a bargain and a great value in the context of the D3x. Which is Nikon's current dilemma and the reason for that low value rating: Nikon needs a D700x (basically a D700 body with the D3x sensor at a price that's within sight of the 5DII).

So, again: best image quality of any current DSLR, but it might not be the right one for you.

So if you're having to read this to figure out if you need to get a D3x, the answer is that you probably don't. Those of you who need the pixels have the clients that'll pay for them have probably already run to a store that has one to test it yourself. You already can guess the answer for you, and I hope that the more detailed results analysis I've done above helps that process. For the rest of you: wait for a D700x and re-evaluate.

Finally, a word about "size." One reason a lot of people get excited about the 24mp number is that they envision "printing bigger." At the 360 dpi that you would normally give an Epson inkjet printer (Canon's and HP's prefer 300 dpi, I believe), the maximum size of a D3x image before you have to start resizing is ~11x17". On a D3, that number would be ~8x12". As I've pointed out elsewhere, stitching on a D3 can get you into D3x realm pretty easily. The D3x has more of a disadvantage against state-of-the-art medium format backs than the D3 has against the D3x, in my opinion. If you're really trying to print big, the world hasn't changed since film: you probably need medium format.

Bottom line: the D3x produces excellent, best-of-class images, but that comes at a price that's the highest of any existing DSLR. You'll need really top notch lenses and technique to resolve that performance. Are you really ready for all that portends?


  • You're paying for pixels. You're paying 3 cents for every 100 pixel gain. That's the highest premium I could find anywhere.
  • Frame rates went backward. If you're trying to get every bit of performance by shooting 14-bits, the frame rates are a real step backwards (1.8 fps), but even a 12-bits this isn't a sequence shooter (5 fps).
  • Missing elements. No sensor cleaning and no video, for example. The tech base of the camera feels a little dated because of that.


  • Moving towards Medium Format Country. Yes, the extra pixels perform. I'm pretty comfortable saying this is the best image quality of any 35mm-sized DSLR.
  • It's a D3. Lots of control, lots of customization, lots of performance, and robust.
  • Surprisingly decent high ISO. Up to ISO 1600 is very usable. And noise tends to be film like with little of that false color thing digital usually adds. | Nikon | Gadgets | Writing | imho | Travel | Privacy statement | Copyright 2009 Thom Hogan. All rights reserved.