[Nicolas06]

I agree. I've turned off the noise reduction in the menu on my KP and it's still pretty good up to ISO 3200. I don't have a particularly high tolerance for noise, and I won't shoot at ISO's that produce so much noise that it smudges detail. With my K-5iis I was comfortable shooting up to ISO 1600. With the KP I'm comfortable shooting up to ISO 3200. While I wouldn't suggest that the KP is a stop better than the K-5iis (or K-3), if you factor in the greater resolution, it can't be that far off from a stop. Per usual, we're talking subjective perceptual evaluation, and YMMV.

The camera may do all it want, including on raws, regardless of the settings you used in the menus. The K5 was adding blur, the K3 didn't or much less and so the K3 was seen are more noisy because basically you add to do it yourself.

I am sure the KP is better at high iso, but the question is how much once the software part is removed. It can't be 1.5EV or more otherwise it would match K1 performance and it doesn't. The raws on the link at high iso look very blury, but if there was no NR applied to it, they would not look like that. Softened and all.

Overall the work on the KP is really nice and ensure everybody benefit of it out of the box and maybe you can 1 more EV out of it, but I think it highly depend of the scene too.

[rawr]

the question is how much once the software part is removed

It's not software. It's hardware for the KP. They now have a dedicated ASIC (ie the 'Accelerator Unit') handling all this.



Interestingly, the KP web site also says about the Accelerator Unit:

The PENTAX KP couples a new-generation CMOS image sensor with the PRIME IV imaging engine and an accelerator unit to expand the upper limit of the sensitivity range without generating annoying noise. With a top sensitivity of ISO 819200, it assures super-high-sensitivity photography to deliver high-resolution, rich-gradation images with faithful reproduction of the subject’s texture and a sense of depth, even at higher sensitivities. It also produces high-quality images across the entire sensitivity range, from the lowest to the highest.

http://www.ricoh-imaging.co.jp/english/products/kp/

So it's not just doing good things for high-ISO. It probably helps out on pixel shifted, long exposure and other tricky image types too.

[ffking]

i have plenty of patience (i.e. No money)

true that

:d :d :d :d :d

(smilies not working for me - yes, I did type upper case D )

[Nicolas06]

It's not software. It's hardware for the KP. They now have a dedicated ASIC (ie the 'Accelerator Unit') handling all this.



Interestingly, the KP web site also says about the Accelerator Unit:



PENTAX KP | RICOH IMAGING

So it's not just doing good things for high-ISO. It probably helps out on pixel shifted, long exposure and other tricky image types too.

Maybe in help for many things but this accelarator unit is just specialized processing power. It is the software that run, the quality of the algorithms that define what is done and the same algorithms could be run on any processing unit with the same results. You could have the same code running on your computer. If done on a GPU, it might be even faster.

The part of their propaganda that speak of a new CMOS sensor is the part that supposedly can't be achieved by a K3 + appropriate software. The part with the accelarator unit can be done on any camera from the raw with the right software. The accelerator unit make it fast and low power enough to run on the camera but that's about it.

---

Last edited by Nicolas06; 11-05-2017 at 03:32 AM.

[Kunzite]

The part of their propaganda that speak of a new CMOS sensor

Propaganda? Do you have any proof it's not a new sensor, then?

I'd say this thread had turned into a very misleading direction, advising people to ignore the image quality advantage of the newer cameras. Because it seemingly can be reproduced with software, but nobody did that - I wonder why.

[Nicolas06]

Propaganda? Do you have any proof it's not a new sensor, then?

I'd say this thread had turned into a very misleading direction, advising people to ignore the image quality advantage of the newer cameras. Because it seemingly can be reproduced with software, but nobody did that - I wonder why.

Propaganda:

information, ideas, or rumors deliberately spread widely to help or harm a person, group, movement, institution, nation, etc.

Propaganda isn't about lying it is about being deliberate. Still from what rawr quoted, the KP certainly do not deliver high quality image accross the sensitivity range. There no camera that do that actually. This link was really propaganda. Not saying it is bad or uncommon, but all companies do that focussing only on the positive aspect, exagerating quite a bit and forgetting to speak of eventual limitations/weaknesses. For me this is propaganda.



There is difference in high iso coming from KP in the way that if you work with software with raw, you can certainly narrow the gap with K3-II but a difference remains, in particular at extremely high iso.

For me this come from the new CMOS sensor or better NR algorithm than available to me (like DxO prime or RawTherapee). In the end it doesn't really matter as to get that performance now you need a KP in APSC.

Doing a comparison from the RAW of dpreview + their jpeg I would say that very roughly:
- KP JPEG are better by rougly 1EV than K3 JPEG
- K3 RAW + DxO prime are better rougly by 1EV than KP JPEG
- KP RAW + DxO prime are better rougly by 1EV than K3 RAW + DxO prime.

Overall KP got 1EV better than K3. You can get it quick with shooting JPEG but then you'll not get really better than just applying DxO prime in batch on your old K3... Actually the results may be worse.

But if you go the full spectrum from K3 JPEG to KP RAW + prime you get 3EV. 2 EV from DxO prime, 1EV from the KP.

But using great NR algorithm to get better results isn't new, So the figure to retain is one more EV for KP than K3.

[rawr]

the same algorithms could be run on any processing unit with the same results

Not really. Some of those image processing chips (like the Milbeauts) have been engineered to perform specific image processing tasks.

Sure, those ASICs run micro-code and algorithms like all computers do, but it's like saying any old general purpose Pentium chip can do the same job as your Nvidia GPU in texture mapping and rendering while playing a graphics intensive game.

And I don't understand the constant, admiring references to DxO Prime. I've used DxO for years, and Prime is still a horribly smudgy, error prone and S - L - O - W NR tool. Not a good example.

[Nicolas06]

Not really. Some of those image processing chips (like the Milbeauts) have been engineered to perform specific image processing tasks.

Sure, those ASICs run micro-code and algorithms like all computers do, but it's like saying any old general purpose Pentium chip can do the same job as your Nvidia GPU in texture mapping and rendering while playing a graphics intensive game.

And I don't understand the constant, admiring references to DxO Prime. I've used DxO for years, and Prime is still a horribly smudgy, error prone and S - L - O - W NR tool. Not a good example.

The image processing chip is still a turring machine, the same work can be done by any turring machine, that's computer science 101. I don't think the accelarator unit is a quantic computer or something like that. And between the Nvidia GPU or the accelerator unit as the one being faster to do the work, I bet on the Nvidia GPU.

For DxO prime, well you may not like it, that's your problem. Still the result look better than the Pentax jpeg that are more "smudgy and error prone" than prime is. You can also learn to use the right setting for your needs with the various options. Or another software NR tool. The basic one on DxO is alredy much better than the Jpeg outbut and the one bundled with lightroom is also quite well performing.

[Kunzite]

Propaganda:
Propaganda isn't about lying it is about being deliberate. Still from what rawr quoted, the KP certainly do not deliver high quality image accross the sensitivity range. There no camera that do that actually. This link was really propaganda. Not saying it is bad or uncommon, but all companies do that focussing only on the positive aspect, exagerating quite a bit and forgetting to speak of eventual limitations/weaknesses. For me this is propaganda.

Then, what you're doing here is also propaganda.
At least you're now admitting there is some difference from the K-3... perhaps later you will try - and fail - to obtain similar images starting from the K-3 RAWs and processed on Nvidia GPUs.

[Nicolas06]

Then, what you're doing here is also propaganda.
At least you're now admitting there is some difference from the K-3... perhaps later you will try - and fail - to obtain similar images starting from the K-3 RAWs and processed on Nvidia GPUs.

And ? What problem do you have with that word propaganda ?

The results you get on DxO could be done on your GPU or CPU. Depend if you activate OpenCL or not. It is worth it if you have a fast GPU. The actual result shall be exactly the same anyway, only the speed would differ.

I don't know if lightroom or Capture One have GPU support but I wouldn't be surprised for it to be the case.

[Kunzite]

For me, it has a negative connotation - but I get it that it doesn't for you.

How about the result you get with the KP's sensor and Pentax own image processing hardware and algorithms? We started from Winder's claim that it's all software, something I don't think you (fully) agree with. Let's be clear what I disagree with:

a K-3 RAW file processed by DxO Prime would give as good or better results than the KP in terms of noise.

I'll ask again, though I know the answer is a no: did anyone believing this claim actually tried it?

I hope we all agree the KP's sensor is a newer generation and not "propaganda".

[Winder]

For me, it has a negative connotation - but I get it that it doesn't for you.

How about the result you get with the KP's sensor and Pentax own image processing hardware and algorithms? We started from Winder's claim that it's all software, something I don't think you (fully) agree with. Let's be clear what I disagree with:

I'll ask again, though I know the answer is a no: did anyone believing this claim actually tried it?

I hope we all agree the KP's sensor is a newer generation and not "propaganda".

I did not say it was all software. I said the camera was applying NR to the image before it was saved as a DNG or RAW file. Which is basically what the Ricoh marketing material says as rawr posted above.

I'm not sure that it is a new generation of Sony sensor. Sony has not introduced a new SKU for a 24MP APS-C sensor. Maybe you can link to the new generation of Sony sensor? Its not the a6500/6300 sensor because that has PDAF on sensor. The IMX193 is what is in all of the 24MP APS-C DSLRs. The IMX 271 is the 24MP sensor with PDAF on chip and that is used in the Fuji X-T2, a6500 and all mirrorless APS-C cameras that have a 24MP Sony APS-C sensor. These are the only Sony 24MP APS-C sensors currently in production. Products Line up for Camera Image Sensor for Camera | Sony Semiconductor Solutions Corporation

Please post the link to the new generation of Sony 24MP APS-C sensor you are referring to.

[normhead]

Sorry I deserted you positive guys. I'm a happier person with all these dudes on ignore.

I should point out, this nonsensical Pentax bashing you all engage practically full time, is not news.

It got old long ago.

Oh wait, there is absolutely no thread that can't be interspersed with some completely unfounded Pentax bashing, which usually follows the usual litany of complaints about Pentax AF etc. etc.

But I have learned some stuff by reading some of the quotes.

The new Pentax Accelerator unit, shown in the photo is marketing hype. It doesn't do anything you can't do on your computer.
Pentax has to catch up with all the other camera companies in AF, but the other camera companies don't have to catch up with Pentax in value for price.
DPR is a reliable resource.
Pentax cannot possibly have thought of a technical way of improving a camera that some other camera company didn't think of first (Pixel Shift was the exception that proves the rule.)

You guys are so full of bright cheery accurate information. More fun than a barrel of monkeys. <dripping with sarcasm.>

I guess no one can, just say "No, no news." That opportunity to bash Pentax and tell everyone how bad they are, just can't be missed.

I see how this works.

[Kunzite]

Winder, I quoted your claim. I'm observing again the complete lack of any proof - or any sign that you actually investigated the issue.
I don't care what you believe; Ricoh Imaging says it's a new sensor so it is. It's not like they were stuck with the same sensor since 2013; remember how they used a 20MP sensor in the K-S1 and K-S2? The PDAF capable 24 MP sensor on the K-70?
Then, there's the different character of the KP's and K-3's sensors. One was known to be noisier than the previous 16MP version since it was launched; the other is one of the bests on the market, today. Take a look at this; I won't believe you can add the missing detail back by doing noise reduction in DXO Prime:
Studio shot comparison: Digital Photography Review

[reh321]

Maybe in help for many things but this accelarator unit is just specialized processing power. It is the software that run, the quality of the algorithms that define what is done and the same algorithms could be run on any processing unit with the same results. You could have the same code running on your computer. If done on a GPU, it might be even faster.

The part of their propaganda that speak of a new CMOS sensor is the part that supposedly can't be achieved by a K3 + appropriate software. The part with the accelarator unit can be done on any camera from the raw with the right software. The accelerator unit make it fast and low power enough to run on the camera but that's about it.

The image processing chip is still a turring machine, the same work can be done by any turring machine, that's computer science 101. I don't think the accelarator unit is a quantic computer or something like that. And between the Nvidia GPU or the accelerator unit as the one being faster to do the work, I bet on the Nvidia GPU.

For DxO prime, well you may not like it, that's your problem. Still the result look better than the Pentax jpeg that are more "smudgy and error prone" than prime is. You can also learn to use the right setting for your needs with the various options. Or another software NR tool. The basic one on DxO is alredy much better than the Jpeg outbut and the one bundled with lightroom is also quite well performing.

I don't understand why all this Computer Science talk is here. I'm guessing this is not a true Turing Machine - it doesn't go back-and-forth on a tape, but instead does calculations very quickly. None of this matters.

What does matter is that all testimony says the K-70 and KP are producing good, high quality images at high ISO values Straight Out Of Camera; yes, people could do some of that PP, but some of us would rather spend hobby time in the field than pressing keys, and they are reaching out to us. I don't care how they do it. It doesn't matter! The K-1 has a much better "sports" DxOMark score than the Nikon D850 does. That is very good news!! Pentax seems to have chosen graceful high ISO performance as a hallmark, something showing up in all their new products; they are doing it, doing it well, and I believe that is very good news.

---------- Post added 11-05-17 at 06:06 PM ----------

Maybe in help for many things but this accelarator unit is just specialized processing power. It is the software that run, the quality of the algorithms that define what is done and the same algorithms could be run on any processing unit with the same results. You could have the same code running on your computer. If done on a GPU, it might be even faster.

The part of their propaganda that speak of a new CMOS sensor is the part that supposedly can't be achieved by a K3 + appropriate software. The part with the accelarator unit can be done on any camera from the raw with the right software. The accelerator unit make it fast and low power enough to run on the camera but that's about it.

It is not "propaganda". It is news! Running fast and low power on the camera is very good news.

Nobody cares how fast it could run on the computer back home - this unit is getting the job done in the field, before they ever get home. Nobody is disputing that.

When Tom, Dick or Harriett returns home and unloads images from camera to computer, there is negligible noise despite having using higher ISO values. Nobody is disputing that.

It is happening in real time. Nobody is disputing that.