[rawr]

When the light is good, some higher ISOs look okay....Noise destroy color deph, dynamic range and subtle tones.

This is of course all true. That's why all those well-lit studio test shots (like the Imaging Resource comparometer ones) reveal very little of real/visible ISO performance. Dimming the lights puts high ISO to the test.

But even dark, poorly lit scenes can render well under high ISO. It all depends on the scene. There are no hard and fast rules. Nice solid blacks and dark shadows can cause little visual problem for colour or noise, even at ISO 6400 on either the K-3 or the K-5 (pics below related). What becomes more important is whether you have any motion blur or missed focus, and get the white balance OK. If focus is good, even the highest ISO's can look decent, perhaps with appropriate post-processing using LR, DxO, C1, even PDCU.

K-3

Dee Dee Lavell opening
Tamron AF 17-50mm F2.8, ƒ/3.2, 50.0 mm, 1/125, ISO 6400

K-5

Guitar and drums
PENTAX-FA 50mm F1.4, ƒ/1.6, 50.0 mm, 1/100, ISO 6400

[zx-m]

Hence, in weak lighting I am perhaps gaining a stop. My D200 looks good at 400. Perhaps I will get 800 out of the K-5ii, but conservatively, probably more like 640. All that for 2/3 of a stop?

[rawr]

All that for 2/3 of a stop?

1 stop or less is the difference between a 24MP full-frame Nikon D610 and a 24MP K-3.

It can make a visible difference sometimes, mainly if you have a sensor like almost any of the new Sony designs (in Nikon, Pentax, Sony) that have very good dynamic range and very 'elastic' RAW's, which allow you to stretch them a LOT via shadow boosting with little image penalty. Pic below related (not mine) - I couldn't do this sort of push with my old 10MP CCD K-200D, even starting from very low-ISO, but the K-5 can do it very cleanly:

[Just1MoreDave]

My understanding is that current sensors gain 1-2 EV compared to what you could expect 4-5 years ago fom say in Pentax land a K7. With good denoising software (DxO, Lightroom), you also make high iso much more pleasing to see than before. 1-2 EV more, depending on the scene.

I just got a K-5 IIs and still have a K-7, so I got curious enough to take some test shots myself. I put one camera on a tripod and set it for a decent histogram. I used the same Pentax-F 28mm f2.8 lens on both cameras, setting focus once and leaving it there. I used the two second delay which disables SR. I set both cameras to RAW, f4.0, 0.5 sec., ISO 6400. I imported the shots into Lightroom 4 with the embedded camera profile for both. I tried to set white balance equally using the same point, not really successful. I exported the shots, downsizing to 800 pixels wide with no sharpening. The K-7 is first.





The big difference here is dynamic range. The K-7 shot has noise, but LR4 can make it look OK. The dynamic range is still pretty bad. I saw a test that said less than 4 stops at ISO 6400.

I got used to the K-7 so I get a nice laugh about K-5 users worrying about ISO 800. I might start thinking that way after several years, but right now I'm enjoying the extra freedom.

[sandilands]

Don't forget my "test" was not a typical shooting situation. Even the "analysis" technique was meant to find noise. In most cases we optimize the shooting situation as best we can to avoid noise. The values on the graph can be a bit frightening when you notice that the RGB values of a pure black space are nearly white (200,200,200) at ISO 51200. This would imply that an image at 51200 would have nearly zero contrast
but we know from experience that even at 51200 although grainy, we can make an image with a recognizable subject. I think the problem with the test technique is that I cannot select a threshold level of less than 1. That means the slightest non black pixel (even if its RGB value is 1, 1, 1) is going to be forced to white (255, 255, 255). In reality the fact that there are still some purely black pixels at 51,200 is remarkable.

So don't get too concerned how the slope increases so severely at ISO 3200 As others have posted, real life photographs can be excellent even at high ISO.

---------- Post added 09-26-2014 at 07:55 AM ----------

And after seeing Dave's car photos (my K-7 experience was similar), I want to design some simple controlled environment test for dynamic range.
I did a luminosity studio test with a friend who has a Canon 5D MKIII. I need to study up on the meaning of dynamic range vs. what I think we tested which I will call static' range.

Basically a continuous light setup and a photovision 3 bar calibration target with pieces of white and black lace stretched over the target to provide some texture. We balanced the lights (Videolux 5600K) as evenly as possible and used the target to get perfect exposures which we both set at f/5.6 ISO 100 for 2 seconds. Then we made shots from 30 seconds to 1/8000 sec. Then we imported all the images into Photoshop, zoomed in, and looked for details in the white and the black lace. Bottom line was the K-5 shots showed details in the black lace over the dark section of the target 3 stops stops faster (darker) than did the 5D. On the other hand, we could detect details in the white lace over the light section of the target 1/3 stop slower (brighter) from the 5D than we did the K-5 .
Conclusion: The K5 has a static range 2.7 stops larger than the Canon 5D MKIII, therefore is a better choice of camera when shooting in caves ... I guess.

I like doing simple repeatable experiments like this. I know the reviewers have all kinds of fancy measurement gear, but these tests can be done in my basement with gear that is commonly available. I would be very interested in anybody else's results doing similar experiments.

Oh and I'll post the graph when I get the chance.

[kh1234567890]

You could possibly save yourself a lot of effort by acquiring a copy or RawDigger.