Friday, November 5, 2010

Pentax K5 Leads DXOMark … but by how much? Examing the K5 versus the Kx versus the D3100 … the newest kids on the block …

And of course we are now seeing bold statements like “there is a a new king” yada yada yada … another poster uses the words “blow away” with respect to the new Pentax lead over others … but I consider that sort of hyperbole more like “blowing smoke” … lots of heat and light, not much fire.

So let’s take a closer look here to see if Pentax are going to wax the competition in this generation of sensors. First, let’s examine the DXOMark comparison of the Kx and the K5 against Nikon’s entry dSLR, the D3100. The D7000 is not available on DXOMark yet, but I’ll look at that one when it arrives …

image

What you see here (click the image to go to the comparison on DXOMark.com) is that the D3100 and the Kx are essentially identical in noise. The Kx does take the lead, building slowly to a full stop at 12800, but frankly, that’s a rather silly level at which to shoot with APS-C. With certain subjects, of course, it can be used … but the image must be fairly dark overall to hide the rampant noise and smoothing that goes on.

Note also that the level at which an image is considered “excellent” is 30dB on this graph … and all three cameras are only above that at 800iso. The Kx has about 1/6 stop lead here, which is effectively invisible, and the K5 has only 1.5dB, or 1/2 stops lead. While it might be considered useful to have that half stop, it certainly is not blowing the D3100 away. Skill with exposure and processing makes far more difference than that difference in RAW SNR.

To be fair, the K5 is only 1/3 stops below 30dB at 1600iso, while the others are a bit further away … so this 2/3 stops difference might be visible under some circumstances.

And after that, the Pentax cams use some form of smoothing that is significant enough that DXO labs felt it necessary to call it out explicitly on their charts. I don’t really know what that means, but I can guess that fine, low-contrast details might be an issue at extreme ISOs. We’ll see as the review pour in I suppose.

The numbers on that graph do show that the Kx is well ahead of the D3100 at 6400iso, and is actually about equal at 12800iso to the D3100 at 6400iso. So how do we see what that means? Well, that’s pretty easy, since imaging-resource.com provides samples from both of these cameras shot under the same lighting. Once we normalize the D3100 to the Pentax size (it;s a tiny amount, but let’s make the images identical in size), we can see how this works out in real life.

image

Well, that’s nothing to write home about …. is it … in fact the D3100 looks about the same as the Pentax at 6400iso, with a bit more noise perhaps (not much) but with a bit less feeling of “helmet hair”, which is very prevalent with the Kx images. I suppose we can all understand what “smoothing” means when looking at these. The 12800iso image is not close at all to the D4100 at 6400iso ,,, much more smoothing and a lot of mottling present there.

So for low light work, any of these cameras is going to ve very suitable. The K5 is definitely leading by the time we reach 1600iso, but not enough to make a massive difference in practical use. Skill will easily trump a 2/3 stops lead. But regardless, if this “smoothing” does not prove to be a trick that costs too much detail, the K5 looks to be a pretty good low light cam.

But let’s take a look at dynamic range before we get too far ahead of ourselves.

imageNow here, Pentax is looking like a miracle worker. Look at the lead at base ISO. Almost 3 full stops! Wow!. There is not much question that one should seriously consider the Kx for landscape and perhaps for wedding use. That’s a non-trivial lead. And here the term “blow away” is pretty appropriate. So with dynamic range, there is definitely some fire …

Note, however, that when you get indoors and start shooting at 800iso and above, the dynamic range difference between the D3100 and the K5 has narrowed to less than one stop at 800iso and about a half a stop at 1600iso. So wedding shooters won’t see much difference (and these cams are in a very different cost class.) The more appropriate comparison is the Kx to the D3100, where the Nikon leads by .2 and .4 stops at 800iso and 1600iso respectively. Not a big lead, but the point being that the dominance of Pentax is not complete. These are very common ISOs at which one shoots inside.

So … landscapers can celebrate that the K5 is their newest answer for 1600 bucks (current Adorama price for the body.) This is pretty decent pricing for such a capable camera. The Kx is actually quite cheap at just over 500 bucks while the D3100 is a bit more money than that … at least 140 more I think. But I also think a comparison of the rest of the features and the systems might prove interesting … anyway, all three are great cams.

I’ll run something similar when the K5 samples are available on imaging-resource.com and when the D7000 has been tested by DXOmark and also has samples on imaging-resource. As with the above low-light detail test, things are not always as expected when the rubber meets the road …

Caveat: The charts are based on measurements in RAW mode … i.e. pretty much what the *sensor* is capable. The actual images compared represent the jpeg engine output, which is what the *camera* is capable of when you let it do the conversions from RAW. Make of that what you will.

5 comments:

Henrik Jansberg said...

I have been looking at DXOmark a couple of times but find the test quite hard to read.
The latest was a graph about light loss in sensor. The author Mark Dubovuy says: "you may get better results with smaller aperture lenses, since they have less light loss". - that is like saying your better of with cheaper lenses?

Kim Letkeman said...

Henrik, I poked around a bit to find the article but I can't see it. Do you have a link?

Henrik Jansberg said...

http://www.luminous-landscape.com/essays/an_open_letter_to_the_major_camera_manufacturers.shtml

Only the graph is DXo the rest is an article on luminous landscape..

I just saw that the D7000 scored an 80 in Dxo

Kim Letkeman said...

Henrik: Yes, I was looking on the site you originally mentioned, which is why I did not find the article :-)

So now that I have read the article, I understand what he is saying. Since the manufacturers "game the system" by boosting ISO to compensate for the loss of light caused by very wide apertures, there might be little reason to spend the huge amounts of money on these expensive lenses.

And he might be right. BUT ... the ISO is only increased by half a stop at f/1.2 through f/1.4 so there is still significant gains ... and modern full frame dSLRs are so good at high ISO that the extra half stop will be almost unnoticeable. At f/2 the difference is 1/5 stops, which really *is* unnoticeable.

As for his assertion that the depth of field advantage might not be real because of this effect, I think his premise is silly. If the effect was unnoticeable on digital SLRs, the lenses would not sell at that price. The effect at f/2 is 80% there, according to the loss so you will see it for sure. The effect at f/1.2 is at least 50% there, so again it will be visible.

People have been making compromises by getting Canon's f/4 series of lenses for years ... they raise the ISO to compensate and live without the extra depth of field. But people who buy the really good lenses know that they are wicked sharp wide open most of the time, so the advantages just pile up ... for a big, big price.


I frankly think this is a tempest in a tea pot.

Henrik Jansberg said...

Does that mean that it is better to spend more on the camera and less on the lenses?

I mean on my old 350d I can very clearly see that shooting f2 at iso 800 is far better (sharper) than f2.8 at iso 1600..