[bdery]

Is there real life difference in AF and Metering btw K3 and K5IIs?

Yep. The K-3 is better in both cases, although for AF it's highly dependant on the lens (and with SDM it's less obvious). Metering is improved, especially with PTTL. Auto WB with multizone is also very effective.

Regarding noise, a lot of things have been said, many of them innacurate. It boils down to this : per pixel, at full resolution, in the same conditions, the K-5 will be slightly better. Because its pixels are larger. It,s physics. For for the whole image (whether you see it on screen or print your image) the K-3 will deliver better results.

Long story short, I can shoot at ISO 5000 without second thoughts.

[Spodeworld]

For for the whole image (whether you see it on screen or print your image) the K-3 will deliver better results.

Can you elaborate on this? Does this include noise?

Thanks

[Rondec]

Can you elaborate on this? Does this include noise?

Thanks

Yes. Assuming the same viewing/printing size, K5 and K3 files will be equivalent, except that K3 files will have more resolution at lower isos.

[bdery]

Can you elaborate on this? Does this include noise?

The whole question of noise comparisons between K-5 and K-3 arises from the fact that the resolution is different. Pixels on the K-3 are smaller. So if you compare noise between two adjacent pixels, the larger pixels of the K-5 will deliver less noise (a basic signal-to-noise ratio matter, no way around it, electronic noise will be mostly constant and signal will be higher).

However it makes little sense to compare the per-pixel performance.

If you compare IMAGES, looking not at two adjacent pixels but at the global result (be it in print orr on screen) then that higher per-pixel noise is averaged because you have more pixels. Then the K-3's resolution gain improves sharpness and thus general IQ.

and with the K-3, you can always choose to shoot at 16 MP in JPEG...

[jsherman999]

...

Regarding noise, a lot of things have been said, many of them innacurate. It boils down to this : per pixel, at full resolution, in the same conditions, the K-5 will be slightly better. Because its pixels are larger. It,s physics. For for the whole image (whether you see it on screen or print your image) the K-3 will deliver better results.

So, to tie in with another thread: based on this IQ alone, would you buy the K5 based on it's 'per-pixel' performance** or the K3 based on it's 'whole image' performance***?

** = DXOmark 'screen' tab
*** = DXOmark 'print' tab

[Des]

So, to tie in with another thread: based on this IQ alone, would you buy the K5 based on it's 'per-pixel' performance** or the K3 based on it's 'whole image' performance***?

** = DXOmark 'screen' tab
*** = DXOmark 'print' tab

I think the answer lies in a comparison between real world images. In the excellent PF review of the K-3 there were detailed comparisons between images from the K-3 and the K-5iis, testing various parameters. Pentax K-3 Review - Introduction | PentaxForums.com Reviews (See particularly pages 14 and following). The K-3 came out clearly ahead:
"The Pentax K-3's new sensor has performed very well throughout our tests. We find that it delivers the best overall image quality of any Pentax DSLR to date when looking at the final product, both from JPEG files as well as RAW files. Because the K-3's sensor lacks an antialiasing filter, its files will appear very sharp out-of-camera when paired with quality lenses, but if moire ever becomes an issue, the camera still gives you two separate tools to reduce or eliminate it without having to rely on desktop applications. No other camera currently offers an on-demand AA filter simulator like the K-3. At minimum ISO, the K-3 is capable of capturing stunning detail that might well rival cameras of a higher class. There is no question that the significant increase in resolution is the first thing that users will notice when looking at the files from this camera. The K-3's white balance is very accurate, and its metering has been tweaked to avoid overexposure: both important improvements over earlier models. At higher ISOs, noise sets in early, but as you saw above, this doesn't prevent the K-3 from matching and surpassing the high-ISO output of its predecessor after post-processing.
The dynamic range of the K-3 may be just shy of that of the K-5 IIs in theory, but in practice, its shadow and highlight detail is still very impressive.
If you often find yourself making large prints, you will surely enjoy the K-3's increased resolution. If you mostly publish web-sized images, then the added megapixels may still come in handy for when you crop. And even if you don't need the higher resolution, the K-3's improved metering and white balance is something that you'll be able to take advantage of every day. Even if we ignored the K-3 's improved speed and other innovative features, we would still recommend it for the sake of the image quality alone."

Low-light high-ISO jpgs from the K-5iis had slightly less noise and more shadow detail than those from the K-3, but the K-3 was ahead if shots were taken in RAW and PP'd a little.

[jsherman999]

I think the answer lies in a comparison between real world images. In the excellent PF review of the K-3 there were detailed comparisons between images from the K-3 and the K-5iis, testing various parameters. Pentax K-3 Review - Introduction | PentaxForums.com Reviews (See particularly pages 14 and following). The K-3 came out clearly ahead:...

Yet the 'screen tab' on DXOMark clearly shows the K5 having better per-pixel performance - how can this be if the actual results of the K3 are the same or better, as you say?

[Des]

I'm sure others will have a technical answer to this question, but personally I'd prefer to be guided by a decent side-by-side comparison of images as an indicator of how the camera actually performs.

[bdery]

So, to tie in with another thread: based on this IQ alone, would you buy the K5 based on it's 'per-pixel' performance** or the K3 based on it's 'whole image' performance***?

Pictures and photographs rarely rely on only two pixels To me the answer is simple, and pretty obvious : both cameras deliver great images, but the K-3 lets the user do things that the K-5 doesn't. The increased resolution, increased speed, more evolved controls, etc, make it a more advanced, and more desirable, camera.

Yet the 'screen tab' on DXOMark clearly shows the K5 having better per-pixel performance - how can this be if the actual results of the K3 are the same or better, as you say?

The answer to that question is written in my post from 2014, above.

I'm sure others will have a technical answer to this question, but personally I'd prefer to be guided by a decent side-by-side comparison of images as an indicator of how the camera actually performs.

Indeed.

[jsherman999]

The answer to that question is written in my post from 2014, above.

So, if someone were to say that the 'screen tab' is the metric to use when judging actual camera output, you'd find that suspect?

[normhead]

IF you bought the K-3 for cropping, I find 640 ISO is the highest I like to go. On K-5 it would be 800 ISO. If I'm using the whole frame, and reducing it to 3000 pixels wide or less, I'll shoot either at 1600 ISO if I have to. You get more noise on a K-3, but you also get more magnification. By the time you eliminate the noise it's pretty much a saw off. Except at 100-400 ISO the K3 noise is not really a factor, then you just get a larger image, with noticeably more resolution.

Those who need better than 1600 ISO should definitely be tracking the performance reviews of the new K-1. It should have something to offer. I'm hoping for 800 on my K-5 being 3200 ISO on the K-1, since there have been some suggestions that 6400 can be clean. If you get some clean images at 6400, usually that means dependably clean at 3200.

On my K-3 I can get clean images at 1600... but sometimes I got noise at 800, dependable at 640 ISO.
On my K-5 I can get clean images at 1600... dependable at 800 ISO.

[ScooterMaxi Jim]

I don't much question the DxO results, but keep in mind that the SNR is based on no NR (or the minimum applied in the circuitry). So, as you get into the higher ISOs you have a bit more sharpness to work with at 24mp as opposed to 16mp to the extent you need to apply NR. Ultimately, the IQ is very close between them. Keep in mind that DxO scores all lenses higher in sharpness with the K-3 than with the cameras that have 16 mp sensors (and computes the varying AA impacts, as well).

The cameras are pretty different in other important ways. The camera that yields a sharper typical result based on AF or flash performance is a far more important than the minor differences in the sensor characteristics.

That said, I certainly do notice that my technique has to be a bit better with the K-3. Even with SR, the camera requires a steadier hand.

[Rondec]

An interesting thread resurrection here. I thought this whole K5 versus K3 subject had been litigated long ago.

It is clear from the graphs that the K5 II and K3 are extremely close with regard to SNR. The differences tend to kick in when the K5 has some noise reduction in the RAW files, which the K3 doesn't have. The K5 II is better with regard to dynamic range at low iso, while the K3 has a little better resolution -- mainly useful in the lower iso segments. The real world implications of all of this are nil. I would choose a K3 over a K5 II 10 out of 10 times -- not because the sensor is better, but because the rest of the camera is better, more nimble and has better features. The only thing the K5 II really has going for it is iso 80, which is handy at times, but not enough to sway the decision in my opinion.

This is iso 3200 on the K3, which is more than adequate for my purposes.

[mee]

An interesting thread resurrection here. I thought this whole K5 versus K3 subject had been litigated long ago.

It is clear from the graphs that the K5 II and K3 are extremely close with regard to SNR. The differences tend to kick in when the K5 has some noise reduction in the RAW files, which the K3 doesn't have. The K5 II is better with regard to dynamic range at low iso, while the K3 has a little better resolution -- mainly useful in the lower iso segments. The real world implications of all of this are nil. I would choose a K3 over a K5 II 10 out of 10 times -- not because the sensor is better, but because the rest of the camera is better, more nimble and has better features. The only thing the K5 II really has going for it is iso 80, which is handy at times, but not enough to sway the decision in my opinion.

This is iso 3200 on the K3, which is more than adequate for my purposes.

The other shot of the tiny bird https://www.flickr.com/photos/85914716@N03/25522005905/ is an even better example of sharpness at high ISO imo. Impressive. What software did you use to render from RAW and did you use any out of camera sharpening?

[bdery]

So, if someone were to say that the 'screen tab' is the metric to use when judging actual camera output, you'd find that suspect?

What do you mean by "screen tab" ?

In any case, when judging a camera, I do NOT rely on DXO, or any other single resource using one single sample unit to draw conclusions for a whole production range. And I do not care for a few percents of variations between two units.

Considering that all current cameras use the same few sensors, I look at what the manufacturer does with that sensor. And more importantly, how the complete tool (the camera body) is designed and thought.