K5: Testing RAW Continuous Shooting

Started Oct 20, 2010 | Discussions
MightyMike Forum Pro • Posts: 38,597
128mb buffer in my estimation

Sounds like the K-5 has a 128mb buffer, that should allow up to 6 or so compressed RAW photos, and at least one more photo to be written within the first second using a 30MB/s card... that would suggest 7 frames before the buffer fills. Possibly 8 low detail frames.

Perhaps its time to buy a class 10 card or SDXC to get the most out of the camera
--
Mike from Canada

'I like to think so far outside the box that it would require a telephoto lens just to see the box!' ~ 'My Quote :)'

http://www.airliners.net/search/photo.search?sort_order=views%20DESC&first_this_page=0&page_limit=180&&emailsearch=mighty_mike88%40hotmail.com&thumbnails=

MightyMike Forum Pro • Posts: 38,597
Re: 128mb buffer in my estimation

running the numbers a little further, with a 30MB/s card it shouldn't take much more then 4 seconds to clear the buffer (not taking into account anything else that might slow it down) and a 20MB/s card should take as much as 7 seconds...

I hope the 9-10 seconds to clear isn't accurate.
--
Mike from Canada

'I like to think so far outside the box that it would require a telephoto lens just to see the box!' ~ 'My Quote :)'

http://www.airliners.net/search/photo.search?sort_order=views%20DESC&first_this_page=0&page_limit=180&&emailsearch=mighty_mike88%40hotmail.com&thumbnails=

MightyMike Forum Pro • Posts: 38,597
Re: 128mb buffer in my estimation

Running the numbers yet even further...

when using AF-C assuming approx. 4fps in shutter priority you can get 8-9 frames in 2 seconds before a 7-10 second (depending on the write speed) buffer empty delay before you are able to do the same burst, it should slow down to approx. 1 shot per second if you keep your finger on the button, however for every second you give the camera between bursts you should be able to burst of approx. that many more frames.

also

High ISO, 4-5 frames, then a 7-10 second (depending on the write speed) delay all cause by the extra size of the RAW files due to noise

I used 7-10 seconds as its close to what is reported despite that ability to possibly be as fast as 4 seconds
--
Mike from Canada

'I like to think so far outside the box that it would require a telephoto lens just to see the box!' ~ 'My Quote :)'

http://www.airliners.net/search/photo.search?sort_order=views%20DESC&first_this_page=0&page_limit=180&&emailsearch=mighty_mike88%40hotmail.com&thumbnails=

MightyMike Forum Pro • Posts: 38,597
Re: 128mb buffer in my estimation

Just thinking about it some more...

there is an obvious reason to keep the buffer down to 128mb, part of the reason is we don't have access to the much faster CF cards. the other part is the buffer empty delay...

a 128mb buffer should empty in at least 5 seconds with a 30MB/s card and 7 seconds with a 20MB/s card...

so you only get 4-8 shots RAW with such a buffer (depending on settings and the images compressibility) that's a bit annoying. But lets say they doubled the buffer size to 256mb, then it would take 10-14 seconds to empty the buffer, sure you'd get a little more then double the images (9-17 or so) but you'd have to wait that much longer, a 512mb buffer would get you even more images (20-36 or so) but that could leave you waiting 30 seconds before another full burst, sure you could burst off a smaller set even sooner but it still leaves you waiting, especially for the review time.

I think since the jpeg quality ain't half bad those who need a larger buffer should consider it.

Don't expect bigger buffers until there is a cheap breakthrough in SD card tech that allows speed up to 120MB/s or higher
--
Mike from Canada

'I like to think so far outside the box that it would require a telephoto lens just to see the box!' ~ 'My Quote :)'

http://www.airliners.net/search/photo.search?sort_order=views%20DESC&first_this_page=0&page_limit=180&&emailsearch=mighty_mike88%40hotmail.com&thumbnails=

Dale108
OP Dale108 Veteran Member • Posts: 7,357
Re: 128mb buffer in my estimation

Hi Mike:

Some good ideas. I haven't checked in to what the Canon 7D or Nikon D300s buffers are like but i would expect them to be larger. If the focus tracking is better then the K7 (haven't tried out in field), then this could work well enough.

Dale
--
http://www.pbase.com/abundant108

http://www.pentaxphotogallery.com/home#section=ARTIST&subSection=272176&subSubSection=1787360&language=EN

 Dale108's gear list:Dale108's gear list
Sony RX10 III Olympus TG-5 Nikon Coolpix P1000 Olympus E-M5 II Pentax K-1 +1 more
dgaies Regular Member • Posts: 290
Re: 128mb buffer in my estimation

I hope the 9-10 seconds to clear isn't accurate.

From the time I start the first shot until the time the orange disk-write light turns off takes exactly 10 seconds (I tried it 3 times). I don't know if that is the correct way to measure the amount of time it takes to clear the buffer, but that's what I was measuring

FWIW, I'm using a 30MB/s Sandisk Extreme III SD card.

MightyMike Forum Pro • Posts: 38,597
Re: 128mb buffer in my estimation

well that just adds to my reasoning as to why the buffer is small, but why does it take approx. 9 seconds to clear 128mb of data from the buffer to the card? especially on a 30MB/s (is that a guaranteed 30? or a max. potential of 30? meaning it could actually be running at 15MB/s for average use. I know the card manufacturer do fudge the numbers. So maybe 30MB/s for a fresh formatted card with no fragmentation. It could be slower the more the card is full. possibly its only guaranteed to be 10MB/s towards the full mark of the card.
--
Mike from Canada

'I like to think so far outside the box that it would require a telephoto lens just to see the box!' ~ 'My Quote :)'

http://www.airliners.net/search/photo.search?sort_order=views%20DESC&first_this_page=0&page_limit=180&&emailsearch=mighty_mike88%40hotmail.com&thumbnails=

MightyMike Forum Pro • Posts: 38,597
Re: 128mb buffer in my estimation

They could afford a larger buffer as they can use 60MB/s CF cards

Dale108 wrote:

Hi Mike:

Some good ideas. I haven't checked in to what the Canon 7D or Nikon D300s buffers are like but i would expect them to be larger. If the focus tracking is better then the K7 (haven't tried out in field), then this could work well enough.

Dale
--
http://www.pbase.com/abundant108

http://www.pentaxphotogallery.com/home#section=ARTIST&subSection=272176&subSubSection=1787360&language=EN

-- hide signature --

Mike from Canada

'I like to think so far outside the box that it would require a telephoto lens just to see the box!' ~ 'My Quote :)'

http://www.airliners.net/search/photo.search?sort_order=views%20DESC&first_this_page=0&page_limit=180&&emailsearch=mighty_mike88%40hotmail.com&thumbnails=

dgaies Regular Member • Posts: 290
Re: 128mb buffer in my estimation

MightyMike wrote:

well that just adds to my reasoning as to why the buffer is small, but why does it take approx. 9 seconds to clear 128mb of data from the buffer to the card? especially on a 30MB/s (is that a guaranteed 30? or a max. potential of 30? meaning it could actually be running at 15MB/s for average use. I know the card manufacturer do fudge the numbers. So maybe 30MB/s for a fresh formatted card with no fragmentation. It could be slower the more the card is full. possibly its only guaranteed to be 10MB/s towards the full mark of the card.
--

I'll try formatting the card in the morning and trying it again. The only other thing I can think of is maybe the buffer clears a few seconds before the light goes out but the light stays on while the camera is processing the images (adding the NR to the RAW files and such).

Ok, I just tried the same thing at low ISO and the light did go out a little quicker (6-7 seconds) so it seems the full 9-10 seconds isn't entirely buffer clearing time.

MightyMike Forum Pro • Posts: 38,597
Re: 128mb buffer in my estimation

The delay for the preview after a full burst may make the write time seem longer also.
--
Mike from Canada

'I like to think so far outside the box that it would require a telephoto lens just to see the box!' ~ 'My Quote :)'

http://www.airliners.net/search/photo.search?sort_order=views%20DESC&first_this_page=0&page_limit=180&&emailsearch=mighty_mike88%40hotmail.com&thumbnails=

dgaies Regular Member • Posts: 290
Re: 128mb buffer in my estimation

MightyMike wrote:

The delay for the preview after a full burst may make the write time seem longer also.

I was only going by when the orange light was on/off and was using a timer, so it was fairly objective.

MightyMike Forum Pro • Posts: 38,597
Re: 128mb buffer in my estimation

was just suggesting another reason for a longer delay then expected
--
Mike from Canada

'I like to think so far outside the box that it would require a telephoto lens just to see the box!' ~ 'My Quote :)'

http://www.airliners.net/search/photo.search?sort_order=views%20DESC&first_this_page=0&page_limit=180&&emailsearch=mighty_mike88%40hotmail.com&thumbnails=

tigerzen Regular Member • Posts: 240
Re: K5: Testing RAW Continuous Shooting

nikon the same...33 shots in reserve under 800 iso...after thant 5 in reserve.

matander Regular Member • Posts: 348
Re: K5 vs K7 AF

Hi,

Would love to see AF performance of K-5 and 50-135, especially for C-AF and focus tracking. I tend to use my K-7 together with 50-135 mostly for moving subjects such as indoor and outdoor (tarck & field) sports. I have not seen any assessment of K-5+50-135+C-AF+predictive tracking yet, so if you find the time i would be very interested in such a test. Acomparison with 60-250 in AF speed would also be appreciated since with K-5´s improved ISO performance the F4 on 60-250 vs F2.8 on 50-135 could be a "non-issue".

Best Regards
Mats

dgaies wrote:

I haven't tried my 50-135 yet, but the 60-250 locks quicker on the K5 as well.

-- hide signature --

K-7, DA*50-135, Sigma 18-502.8 macro, AF540.

 matander's gear list:matander's gear list
Samsung NX200 Samsung NX1 Samsung NX 18-55mm F3.5-5.6 OIS Samsung NX 50-200mm F4-5.6 OIS
k2park Forum Member • Posts: 59
Re: DNG or PEF?

Dale108 ? Have you tried with PEF? PEF must be compressed one, could be better.
Could you confirm it?

GordonBGood Veteran Member • Posts: 6,308
Re: 128mb buffer in my estimation...

MightyMike wrote:

well that just adds to my reasoning as to why the buffer is small, but why does it take approx. 9 seconds to clear 128mb of data from the buffer to the card? especially on a 30MB/s (is that a guaranteed 30? or a max. potential of 30? meaning it could actually be running at 15MB/s for average use. I know the card manufacturer do fudge the numbers. So maybe 30MB/s for a fresh formatted card with no fragmentation. It could be slower the more the card is full. possibly its only guaranteed to be 10MB/s towards the full mark of the card.

Mike, as to why it takes 9 to 10 seconds to clear the size of buffer, whatever it might be, it's because that's how long it took with the K-7 and the basic electronics are the same along with the buffer size. It seems that no matter how fast the flash memory you feed the camera, the write speed stays at something about 15 MBytes per second. Think about this: although the K-7 was writing out about 12 files in those 12 seconds where as the K-5 is writing out only about seven or eight files, they are about that percentage bigger so the amount of data being written is about the same.

You might be right about the 128 MByte raw buffer estimation, as that would make sense in that about four 35 MByte raw files would fill it, with the first file mostly written out to flash memory, but I am starting to take an alternate approach, as follows:

So let's see how your proposal works in terms of the K-7 with it's raw files that averages perhaps 14.5 MBytes or a little more for only lowest ISO shooting: If one can shoot for about three seconds continuously, about three images would have been already written out to memory, meaning that the buffer should contain about 12 low ISO (and maybe quite low detail) raw images when it fills up. This would require that each compressed raw image only be about 10 MP per raw file for the 14.2 MP sensor, which I find likely pretty low for anything more than a black frame image. This also doesn't really make sense in the camera's ability to reprocess JPEG's for raw's that are still in memory: this requires that the raw data be in an unpacked/unprocessed state to be able to feed it to the JPEG processor.

Let's take another approach and assume that (at least for the K-7) that the buffer actually contains pure raw data at two bytes per photosite as well as another buffer sufficient in size to hold the maximum compressed data . Now, the buffer size needs to be about 30 MBytes per image plus another about 22.5 MBytes to hold the compressed data, so roughly 640 MBytes are required to hold the raw data for the remaining about 12 images when the buffer fills up in continuous raw mode.

So, let's switch back to the K-5 and consider that it likely works the same way. It would require about 32.8 MBytes to hold the raw data plus another up to 32.8 MBytes to hold the compressed data for the new 14-bit format. You could fit only about nine of those if the buffer were exactly the same. However, although the physical memory is likely exactly the same size, various buffers used in developing the JPEG preview are certainly larger, and it is likely that the available raw memory has been further reduced by several 10's of MBytes, thus the best case buffer depth of about eight raw images.

Now, this doesn't explain why the buffer depth drops even further for high ISO shooting. I suspect we may well find this is a bug in firmware (although Pentax may never admit that) where the firmware is reserving and not redeeming the space reserved for the compressed data even when it isn't used because it would be larger than the uncompressed data size. I would expect this to get fixed in upcoming firmware updates. Alternatively, perhaps there is indeed even higher Noise Reduction (NR) automatically applied for high ISO raw data and or JPEG previews that is taking extra space. If the regular raw compulsory NR that is used is causing this, I would expect the effect to start at ISO 3200. Does the camera still do this even when all JPEG NR has been turned off?

It would be interesting to confirm this drop in buffer depth at high ISO (and at what high ISO) with everything possible turned off and for the very smallest in file sizes as for a lens cap shot, both to confirm whether the drop in concurrent with a certain file size/detail level and whether the buffer depth is dependent on output file size or not.

Regards, GordonBGood

GordonBGood Veteran Member • Posts: 6,308
Re: DNG or PEF?

k2park wrote:

Dale108 ? Have you tried with PEF? PEF must be compressed one, could be better.
Could you confirm it?

I doubt that PEF is any different than DNG as since the K-7 (K-7, K-x, K-r, and K-5), DNG's have been compressed by almost the same algorithm as PEF's.

Regards, GordonBGood

k2park Forum Member • Posts: 59
Re: DNG or PEF?

Really, I'm K20D user and did not know DNG is also compressed in later models.
In K20D, only pef is compressed.

Freestyler Regular Member • Posts: 170
Re: 128mb buffer in my estimation...

I believe the reason would be cost of the buffer rather than an issue limiting it. 128Mb is what i had thought it would be, the cost per unit to increase that to 256Mb, (especially on a model that is using the same parts as it's predecessor) would be quite large.

Twice as much Nram (or similar) is quite costly at a manufacturing level, Pentax would only do it if they really felt they had to, and on a model where they now buy Sony instead of Samsung sensors, I doubt Hoya would want to incur that kind of any additional prodution costs or profit loss.

But I'm sure future generations using SDXC will implement a larger buffer.
--
http://www.flickr.com/photos/autisticreaction/

Gerry Winterbourne Forum Pro • Posts: 14,474
PEF slower?

k2park wrote:

Dale108 ? Have you tried with PEF? PEF must be compressed one, could be better.

On my K10 I started using PEFs because they were compressed and DNGs weren't, so I could get more on a card. Then I got fed up with the separate XMP files and switched to DNG. I noticed that write time was quicker: I don't know for certain but I've always put that down to the extra time to compress PEFs.

-- hide signature --

Gerry

First camera 1953, first Pentax 1983, first DSLR 2006
http://www.pbase.com/gerrywinterbourne

Keyboard shortcuts:
FForum MMy threads