[BruceBanner]

I know that there is a limit to write speeds, and that using cards faster than 95mb/s seems pointless, but I think I read somewhere that faster cards can actually be important by significantly recovering the camera and clearing the buffer faster? Is there any truth to this?

Has any actual reputable testing been done on this matter?

TIA

Bruce

[Adam]

I know that there is a limit to write speeds, and that using cards faster than 95mb/s seems pointless, but I think I read somewhere that faster cards can actually be important by significantly recovering the camera and clearing the buffer faster? Is there any truth to this?

Has any actual reputable testing been done on this matter?

TIA

Bruce

Yes, this can certainly be the case, but most noticeably with cards with a different speed class. The buffer may clear faster and may or may not result in more frames. Here's a comparison sandisk's extreme with the extreme pro (at the time):

Pentax K-3 Review - Performance and Burst Mode | PentaxForums.com Reviews

It should be noted that the camera generally only locks up to carry out processing (such as distortion corrections), so a faster card will not necessary help with this, but rather the length of bursts and total wait time.

[disconnekt]

Also clears faster when off loading to a computer when used with a good quality (higherish-end) card reader

[sqrrl]

When the media operates faster than the camera can write - the camera becomes the bottleneck.
Faster cards, as in the extreme speed cards are more oriented to video where the camera is taking 30 8mp images a second endlessly, in Pentax cameras the bottleneck is definitely going to be the graphics processing circuitry, which isn't going to run faster due to a card having higher specs than the camera can use.

A fast card is better than a slow card, but you're just going to be wasting money buying cards that write much over the factory specs. Thus 95mbps (I'm just going with your figure here) is a good trade off of speed vs price - if you buy faster cards you get nothing more back for your money, if you buy too much slower the card can't keep up with the camera when shooting at full speed.

As for studies, I don't know, but I'd assume that camera reviewers would probably (each) have a standard test which they would use to test manufacturer claims.

[BruceBanner]

When the media operates faster than the camera can write - the camera becomes the bottleneck.
Faster cards, as in the extreme speed cards are more oriented to video where the camera is taking 30 8mp images a second endlessly, in Pentax cameras the bottleneck is definitely going to be the graphics processing circuitry, which isn't going to run faster due to a card having higher specs than the camera can use.

A fast card is better than a slow card, but you're just going to be wasting money buying cards that write much over the factory specs. Thus 95mbps (I'm just going with your figure here) is a good trade off of speed vs price - if you buy faster cards you get nothing more back for your money, if you buy too much slower the card can't keep up with the camera when shooting at full speed.

As for studies, I don't know, but I'd assume that camera reviewers would probably (each) have a standard test which they would use to test manufacturer claims.

Well... I can understand some of that. I was under the impression someone bought a Lexar 300mb/s card, and although experiencing the same buffer limits and write speeds as a 95mb/s card they did report that once the camera is 'locked up' and experiencing buffer issues that the 300mb/s card would clear that hurdle far quicker and let the user get back to shooting shots again vs the 95mb/s waiting times.
That's what I really want to know. I can also appreciate faster camera to PC transfer times, so I would consider a 300mb/s card if the actual 'clearing the buffer' benefit is there, and if so by how much difference? That's what I would like to know.

[sqrrl]

There is a possibility, but I do suspect that Pentax would use the fastest card available in order to get shot to shot speeds up as far as possible - I haven't noticed a difference between different fast cards at the lower frame rates (ie raw slow) where a faster speed clearing the buffer would make a difference to the number of shots you can take. That's why I am saying that as far as i know the manufacturer quoted specs are likely as fast as they can go.

[BruceBanner]

There is a possibility, but I do suspect that Pentax would use the fastest card available in order to get shot to shot speeds up as far as possible - I haven't noticed a difference between different fast cards at the lower frame rates (ie raw slow) where a faster speed clearing the buffer would make a difference to the number of shots you can take. That's why I am saying that as far as i know the manufacturer quoted specs are likely as fast as they can go.

But this doesn't address the waiting time for clearing the buffer, which is the point of this thread. When Pentax release specs about their cameras they're not going to go that much into detail about it all, just how many burst is possible before reaching buffer. All I know is there I have read user reporting that faster cards can clear that buffer quicker and get the user back to taking shots again, a pretty serious claim if true.
I'm wondering if anyone has done some testing to prove it, a youtube video or something <goes quickly to google youtube pentax buffer videos >

[stevebrot]

When the media operates faster than the camera can write - the camera becomes the bottleneck.

Yep!

Faster cards, as in the extreme speed cards are more oriented to video

Yep!

A fast card is better than a slow card, but you're just going to be wasting money buying cards that write much over the factory specs.

Yep!


Steve

[sqrrl]

Bruce - The buffer is very fast memory which is used to capture raw data from the cameras sensor for processing, it's like a bucket for photographs which then goes through a hose which is the image processing which takes the raw sensor data and converts it into a file which is the raw file or the JPG. It is then saved to the removable user memory, which in this case is the SD card.

In that metaphor
The bucket starts emptying through the image processing hose whenever there is water in it, and it keeps emptying till it is empty - if the bucket is full then you cannot add more - the memory card in this metaphor is like joining on another tube onto the image processing hose - if it's too narrow it causes back pressure to back up the image processing hose - however adding a wider tube which could pass more water than the image processing hose won't speed it up - the sd card tube is passive, it can't suck.

If the fast memory in the buffer and the user memory in the sd card can both run faster than the image processor can process the raw sensor data into image files then the image processor will be the bottleneck in the process.

The burst before buffer in slow frame rates vs fast is a useful measure as the image processor starts working when the first image is taken, it doesn't wait till the buffer is full - you are able to work out from that how much data (the size of the image) can be processed in how long (the time it takes to process an image) and from that you can work out the throughput of the image processor.

If you look at a phantom high speed camera they don't actually do the image processing in real time, instead they burn the raw sensor data directly to ultra high speed memory in the cartridge, processing those into files takes far far longer than actual capture does, and it's a separate process which the user instigates after the fact - that's how they can get so fast, they skip the image processing step and use a massive buffer which is just dumped to super high speed magazines.

[stevebrot]

But this doesn't address the waiting time for clearing the buffer, which is the point of this thread.

'Tis simple...

The time to write is the slower of either the write speed of the card or the controller that marshals the data to the writer or the data bus that delivers the data or the process that flushes the buffer. It is pretty much guaranteed that any of the latter are slower than the current generation of card and that none are optimized to work with that specific brand or model of card.

While it is not reasonable to create a multi-dimensional matrix of all cameras at all combinations of data mix and all cards at all performance spec, Imaging Resource does publish comparative performance figures that might be useful for purchase decisions between contemporaneous cameras.

Your KP:

Pentax KP Review - Performance

Your K-1ii:

Pentax K-1 II Review - Performance


I would not expect either to do significantly better than their testing based on card choice alone.


Steve

[stevebrot]

In that metaphor
The bucket starts emptying through the image processing hose whenever there is water in it, and it keeps emptying till it is empty - if the bucket is full then you cannot add more - the memory card in this metaphor is like joining on another tube onto the image processing hose - if it's too narrow it causes back pressure to back up the image processing hose - however adding a wider tube which could pass more water than the image processing hose won't speed it up - the sd card tube is passive, it can't suck.

...assuming of course that the buffering is relatively simple and without dynamic control or intelligent memory allocation. This all tends to fall into the realm of things we best not dwell on lest they drive us mad.

The practical rule is to buy the best premium performance cards one can afford and to expect performance to NOT improve should we decide to sacrifice to buy the next generation when it becomes available.


Steve

[sqrrl]

Yes, my metaphorical bucket doesn't have intelligent memory allocation.

[biz-engineer]

What has also an effect on buffer clearing time is if the card is crippled by frequent write/cycles in some areas of the memory plan. When I tested the write speed into SD cards, some cards would exhibit a slow down, depending on what zone of the memory is being written, due to the SD controller getting errors during quality checked.

Probably too obvious to mention, but I'll do so anyway: the advertised speed of the card represents the maximum speed at which the card can be read. But, if you take a half-dozen cards from different manufacturers that have the same advertised speed, and you will probably find some whose write-speed is adequate, some that are pretty good, and some will be pretty dismal (e.g., Microcenter's store brand). The read-speed doesn't have any relationship to the write-speed.

As to the camera as the bottleneck, I say that's a good thing. The bottleneck's going to be somewhere, and I don't want the camera to be waiting on other stuff to do its job - I want the other stuff to be waiting on the camera, which ought to be free to do as well as it can.

[BruceBanner]

So, I'm sure I've heard from a number of people say that the new Lexar 300mb/s cards clear the buffer quicker, but of course no difference in fps or when you encounter the buffer, it just recovers quicker. Are most of you saying this is likely placebo or impossible, because the sciencey stuff you know I don't follow very well.

I'm also curious if there is any benefit from writing to lower capacity cards vs higher? Like... I wouldn't expect so... but writing to 4 or 8gb cards (same speeds) to their 64-128gb sized cousins, no difference at all in that respect?