Buffer vs write speed a 30 second test

Started Feb 24, 2013 | Discussions
Stusteelhead Senior Member • Posts: 1,135
Buffer vs write speed a 30 second test
1

Did a quick test with my D300s that I posted in another thread but felt the results deserved there own thread and possibly new testing method for dpreview and other sites to give us the true performance capabilities of these cameras.   The basic premise, how many shots can a camera take in 30 seconds,  6fps means nothing to me if the buffer if always full, so here's what i did,  Set my D300s to 12 Bit lossless compressed Raw, turned off all in camera settings (noise reduction, actice d-lighting etc) then first using a 32GB Sandisk Extreme UDMA 60 mb/s card (fast but not the fastest) and basically started the timer and shot away for 30 seconds to see how many shots my camera could take.  Then I repeated the test writing to the SD slot instead of the compact Flash,  that card was a Lexar Professional SDHC1 16GB 400x Speed.  The results were dramatic and I would love for someone with a D7000 to post the results from their camera.  So with the Compact Flash Card my camera was able to take 83 shots in 30 seconds and with the SD only 46.  I have heard the SD card write speeds were crippled on the D300s and really notice the drop in speed in the field but was still very surprised to see  almost half the shots over 30 seconds.  To me this test should be an essential figure for any new camera specs and means alot more then high fps numbers with small buffers it also speaks to the potential of the new D7100 if it has great write speeds as Jim had mentioned in a previous thread.  Would be very interesting to see the results of other cameras out there using this test.

-- hide signature --

www.raincoastphoto.com

Nikon D300S Nikon D7000 Nikon D7100
If you believe there are incorrect tags, please send us this post using our feedback form.
jfriend00 Forum Pro • Posts: 11,341
Re: Buffer vs write speed a 30 second test

Stusteelhead wrote:

Did a quick test with my D300s that I posted in another thread but felt the results deserved there own thread and possibly new testing method for dpreview and other sites to give us the true performance capabilities of these cameras. The basic premise, how many shots can a camera take in 30 seconds, 6fps means nothing to me if the buffer if always full, so here's what i did, Set my D300s to 12 Bit lossless compressed Raw, turned off all in camera settings (noise reduction, actice d-lighting etc) then first using a 32GB Sandisk Extreme UDMA 60 mb/s card (fast but not the fastest) and basically started the timer and shot away for 30 seconds to see how many shots my camera could take. Then I repeated the test writing to the SD slot instead of the compact Flash, that card was a Lexar Professional SDHC1 16GB 400x Speed. The results were dramatic and I would love for someone with a D7000 to post the results from their camera. So with the Compact Flash Card my camera was able to take 83 shots in 30 seconds and with the SD only 46. I have heard the SD card write speeds were crippled on the D300s and really notice the drop in speed in the field but was still very surprised to see almost half the shots over 30 seconds. To me this test should be an essential figure for any new camera specs and means alot more then high fps numbers with small buffers it also speaks to the potential of the new D7100 if it has great write speeds as Jim had mentioned in a previous thread. Would be very interesting to see the results of other cameras out there using this test.

I like the general direction of your test, but I'm more interested in how many shots a camera can take in a 3-5 second burst and then how long until the buffer is clear again?  I never take 30 seconds of continuous shots, but I frequently take multiple 3-5 second bursts and I'd like to know how many shots the camera takes in that time period.  The 30 second burst is going to fill up the buffer no matter what so it's going to heavily weight card write speed and underweight buffer size and effective fps in my opinion.

-- hide signature --
OP Stusteelhead Senior Member • Posts: 1,135
Re: Buffer vs write speed a 30 second test

Well this was done in bursts, so like I shoot in the field,  I would pause between bursts when I hit the buffer and then at 28 seconds I held down the trigger till it hit 30 seconds with the buffer full.  Basically this test to me tells me whether I will get the shot or not when an owl flies in to grab a mouse kills it then flies off.  Knowing i can take 16 -  5 shot bursts in 30 seconds is rather important to my shooting style and speaks to the overall performance of the camera imo

-- hide signature --

www.raincoastphoto.com

jfriend00 Forum Pro • Posts: 11,341
Re: Buffer vs write speed a 30 second test
1

Stusteelhead wrote:

Well this was done in bursts, so like I shoot in the field, I would pause between bursts when I hit the buffer and then at 28 seconds I held down the trigger till it hit 30 seconds with the buffer full. Basically this test to me tells me whether I will get the shot or not when an owl flies in to grab a mouse kills it then flies off. Knowing i can take 16 - 5 shot bursts in 30 seconds is rather important to my shooting style and speaks to the overall performance of the camera imo

OK, you're more after the same kind of thing I am, but you didn't quite describe it that way in your original post.  I'm very interested in how many uninterrupted shots I get at max fps which your first post didn't really describe (this is mostly dominated by buffer size, but may get a few additional shots due to write speed during the sequence).  I hate the stutter when the buffer fills.  And, then after I've taken a long sequence, I'd like to know how quickly the camera is ready for another long sequence (this is where the real-world card write speed really matters).

-- hide signature --
apaflo Veteran Member • Posts: 3,854
Re: Buffer vs write speed a 30 second test

jfriend00 wrote:

I'm very interested in how many uninterrupted shots I get at max fps which your first post didn't really describe (this is mostly dominated by buffer size, but may get a few additional shots due to write speed during the sequence). I hate the stutter when the buffer fills. And, then after I've taken a long sequence, I'd like to know how quickly the camera is ready for another long sequence (this is where the real-world card write speed really matters).

Nikon provides ballpark numbers in the camera's user manual for buffer size with different formats.

How fast the camera can recover depends on the write speed of the memory card, so that is not possible to specify in any general way.

OP Stusteelhead Senior Member • Posts: 1,135
Re: Buffer vs write speed a 30 second test
1

I understand that, so thats why I put it out to the reviewers to give an idea using a couple of different cards, it only takes 30 seconds and I feel it means more then some numbers on a sheet that never really add up as expected anyway.

-- hide signature --

--

www.raincoastphoto.com

rubank Contributing Member • Posts: 925
Re: Buffer vs write speed a 30 second test

To get a fair (I think) measure of what the D7100 will be able to do, I did test write speed on my D800. If you set the camera to video LV and take stills you end up with 25,3 MP images.

With a really fast SD card (SanDisk Extreme Pro rated at 95 Mb/sec) the D800 writes just about 1 25,3 MP image/sec. That is, for a 6 pic burst it writes for about 6 sec. This is also approximately the write time I get with that card in the USB 3 cardslot of my computer when copying a few files to the card.

So I expect the D7100, which uses the same Expeed 3 as in the D800, to make a halt of around 5 seconds after a 1 second burst (6 14 bit images).

I think that will be an issue for many burst shooters.

jfriend00 Forum Pro • Posts: 11,341
Re: Buffer vs write speed a 30 second test

apaflo wrote:

jfriend00 wrote:

I'm very interested in how many uninterrupted shots I get at max fps which your first post didn't really describe (this is mostly dominated by buffer size, but may get a few additional shots due to write speed during the sequence). I hate the stutter when the buffer fills. And, then after I've taken a long sequence, I'd like to know how quickly the camera is ready for another long sequence (this is where the real-world card write speed really matters).

Nikon provides ballpark numbers in the camera's user manual for buffer size with different formats.

How fast the camera can recover depends on the write speed of the memory card, so that is not possible to specify in any general way.

The point of doing measurements with real world memory cards is to get some real world data.  Who knows where the manufacturers data comes from and/or whether one actually achieves that in the field.  It may sometimes be based on ideal conditions, not real world conditions.  I'd love to hear actual test results with real world cards.  That's what reviews are for.  So, personally I'm not interested only in data that is independent of the memory card or assumes an ideal/perfect memory card.  I want real world measurements.

-- hide signature --
apaflo Veteran Member • Posts: 3,854
Re: Buffer vs write speed a 30 second test

rubank wrote:

To get a fair (I think) measure of what the D7100 will be able to do, I did test write speed on my D800. If you set the camera to video LV and take stills you end up with 25,3 MP images.

With a really fast SD card (SanDisk Extreme Pro rated at 95 Mb/sec) the D800 writes just about 1 25,3 MP image/sec. That is, for a 6 pic burst it writes for about 6 sec. This is also approximately the write time I get with that card in the USB 3 cardslot of my computer when copying a few files to the card.

So I expect the D7100, which uses the same Expeed 3 as in the D800, to make a halt of around 5 seconds after a 1 second burst (6 14 bit images).

I think that will be an issue for many burst shooters.

Probably a good estimate for the D7100 when using that particular SD card.

The D800 can be extremely annoying with buffer overflows for anyone used to faster models.  A D4 zips along for something like 50 shots, shooting RAW, before it hesitates, and even then it just slows down a bit.  A D3S can only do 15 shot bursts.

MMuddler Veteran Member • Posts: 4,526
Re: Buffer vs write speed a 30 second test
2

jfriend00 wrote:

Stusteelhead wrote:

Well this was done in bursts, so like I shoot in the field, I would pause between bursts when I hit the buffer and then at 28 seconds I held down the trigger till it hit 30 seconds with the buffer full. Basically this test to me tells me whether I will get the shot or not when an owl flies in to grab a mouse kills it then flies off. Knowing i can take 16 - 5 shot bursts in 30 seconds is rather important to my shooting style and speaks to the overall performance of the camera imo

OK, you're more after the same kind of thing I am, but you didn't quite describe it that way in your original post. I'm very interested in how many uninterrupted shots I get at max fps which your first post didn't really describe (this is mostly dominated by buffer size, but may get a few additional shots due to write speed during the sequence). I hate the stutter when the buffer fills. And, then after I've taken a long sequence, I'd like to know how quickly the camera is ready for another long sequence (this is where the real-world card write speed really matters).

I'm in the same boat... what counts for me is a) how long a burst I can shoot until it starts to hesitate (and seriously compromise the shot sequence), and b) how long it takes to completely clear the buffer and be ready for another long burst.

In real life I may not fill the buffer on my first burst, but may do so on the second if it is shot right away.... but then I want to know how long it'll be until I can shoot again without the burst stuttering on just the  third or fourth frame. The test should be 1) starting with empty buffer, how many shots at 8fps until it starts to stutter, and then b) how long until all is written to the card. You can figure out your shooting options from there.

-- hide signature --
OP Stusteelhead Senior Member • Posts: 1,135
Re: Buffer vs write speed a 30 second test

There are many ways to do this test obviously but my point is, as a wildlife photographer you often have a very limited time to shoot a bird and I want to know how many shots a given camera can get off in a  small time frame. So this test means  I can get 8 - 10 shots bursts in 30 seconds with 2-4 seconds in between, thats pretty meaningful to me no matter how you break down the shooting sequences.  By comparison the same CF card in my D200 only manages 45 shots the same as the SD card in the D300s.  I would love a D7000 owner to try this and let us know how that camera compares.

-- hide signature --

www.raincoastphoto.com

jfriend00 Forum Pro • Posts: 11,341
Re: Buffer vs write speed a 30 second test
2

Stusteelhead wrote:

There are many ways to do this test obviously but my point is, as a wildlife photographer you often have a very limited time to shoot a bird and I want to know how many shots a given camera can get off in a small time frame. So this test means I can get 8 - 10 shots bursts in 30 seconds with 2-4 seconds in between, thats pretty meaningful to me no matter how you break down the shooting sequences. By comparison the same CF card in my D200 only manages 45 shots the same as the SD card in the D300s. I would love a D7000 owner to try this and let us know how that camera compares.

What you are proposing just isn't a scientifically reproducible test.  You take a burst of unspecified length.  You pause for an unspecified or unmeasured time.  You repeat that process N times.  Sure, this might be what happens in the field, but you can't compare the results of different cameras until you have a reproducible test.  So, you'd either have to drive the shutter by computer to a prepgorammed shutter press pattern or separate out the parts of the process into things that are much easier to measure such as how long to fill the buffer, then how long to clear the buffer. Once you have the latter two numbers as separate values, you should be able to calculate how any other shooting sequence would work.

-- hide signature --
OP Stusteelhead Senior Member • Posts: 1,135
Re: Buffer vs write speed a 30 second test

Fair enough so a repeatable test of this, is just hold the trigger down for 30 seconds and see how many shots.  With the CF I got 100 shots and with SD only 52.  So with an even faster card we are possibly getting to the point where the buffer size is less of an issue, where the camera is able to write to the card almost as fast as the buffer.  Also demonstrates the need for a fast card in the first place.

-- hide signature --

www.raincoastphoto.com

OP Stusteelhead Senior Member • Posts: 1,135
Re: Buffer vs write speed a 30 second test

and the same CF card in my d200 managed 47 shots using the hold down the trigger for 30 seconds method, which clearly shows the performance improvements of the D300s.

-- hide signature --

www.raincoastphoto.com

jfriend00 Forum Pro • Posts: 11,341
Re: Buffer vs write speed a 30 second test
3

Stusteelhead wrote:

Fair enough so a repeatable test of this, is just hold the trigger down for 30 seconds and see how many shots. With the CF I got 100 shots and with SD only 52. So with an even faster card we are possibly getting to the point where the buffer size is less of an issue, where the camera is able to write to the card almost as fast as the buffer. Also demonstrates the need for a fast card in the first place.

OK, that test is repeatable, but it isn't what I care about. If that's what you care about, then we can just agree to be interested in different things.

In your test a 6fps camera that can write to the card at x MB/s and has a 10 image buffer will probably score about the same as a 10fps camera (with the same buffer) that can write to the card at the same x MB/s.  Your test is almost entirely a test of the card write speed since as soon as the buffer fills up, all you can take (regardless of what fps the camera can do) is based on the card write speed.

I'm interested in how many consecutive shots I can take without any stutter (because the buffer filled up) and how fast those shots come (the real fps) because that's really the sequence I want. Once it starts stuttering, you end up missing all sorts of shots you want (at least that's what happens on my D300). Then, once I've done one of those max sequences, I want to know how long until the camera is ready for another max sequence.

-- hide signature --
OP Stusteelhead Senior Member • Posts: 1,135
Re: Buffer vs write speed a 30 second test

I think you are missing my point a bit, sports and wildlife definitely have similar but different needs that I understand.  My point is a big buffer 25 plus shots and fast write speed allow you to shoot a lot of bursts over a short amount of time. no i don't just hold the trigger down for 30 seconds but what often happens is that a subject will fly in for only 30 seconds or less so the most shots I can get over that sort of time span is very important to me.  The fact that the D300 is able to move that much data in 30 seconds means you rarely have to wait on the buffer, and therefore hopefully don't miss the shot.  In sports you will often shoot longer bursts nearly filling the buffer over an action sequence, and need it to empty for the next, I get that, but speed is speed so I feel this test is important to both of us in that its representative of a cameras ability in moving data, like i said i would be curious to see how a D7000 does and the new D7100 with a fast card.

-- hide signature --

www.raincoastphoto.com

jfriend00 Forum Pro • Posts: 11,341
Re: Buffer vs write speed a 30 second test
1

Stusteelhead wrote:

I think you are missing my point a bit, sports and wildlife definitely have similar but different needs that I understand. My point is a big buffer 25 plus shots and fast write speed allow you to shoot a lot of bursts over a short amount of time. no i don't just hold the trigger down for 30 seconds but what often happens is that a subject will fly in for only 30 seconds or less so the most shots I can get over that sort of time span is very important to me. The fact that the D300 is able to move that much data in 30 seconds means you rarely have to wait on the buffer, and therefore hopefully don't miss the shot. In sports you will often shoot longer bursts nearly filling the buffer over an action sequence, and need it to empty for the next, I get that, but speed is speed so I feel this test is important to both of us in that its representative of a cameras ability in moving data, like i said i would be curious to see how a D7000 does and the new D7100 with a fast card.

We can agree to disagree then.  I'm interested in a different characteristic than you are.

I NEVER hold the shutter down any more once the buffer is full and the camera starts to stutter because I can't make sure I'll get the shot I most want or the camera will be ready when I need it to be (if something else starts to happen).  So, I never shoot the way your test runs and thus am not really interested in that measure.

Further, do you realize that your test is really only a measure of card write speed and doesn't show any significant difference between a 10fps camera and a 5fps camera and will barely show the difference between a small buffer and a medium size buffer.  I actually think both of those matter a lot to you too and they really matter to me.  Anyway, no point in further discussing - I was just trying to explain what your test was missing, but if that's what you care about, that's fine with me - just not what I'm interested in.

-- hide signature --
pavi1 Veteran Member • Posts: 6,852
Re: Buffer vs write speed a 30 second test

jfriend00 wrote:

Stusteelhead wrote:

I think you are missing my point a bit, sports and wildlife definitely have similar but different needs that I understand. My point is a big buffer 25 plus shots and fast write speed allow you to shoot a lot of bursts over a short amount of time. no i don't just hold the trigger down for 30 seconds but what often happens is that a subject will fly in for only 30 seconds or less so the most shots I can get over that sort of time span is very important to me. The fact that the D300 is able to move that much data in 30 seconds means you rarely have to wait on the buffer, and therefore hopefully don't miss the shot. In sports you will often shoot longer bursts nearly filling the buffer over an action sequence, and need it to empty for the next, I get that, but speed is speed so I feel this test is important to both of us in that its representative of a cameras ability in moving data, like i said i would be curious to see how a D7000 does and the new D7100 with a fast card.

We can agree to disagree then. I'm interested in a different characteristic than you are.

Perhaps you could share your numbers with us.

-- hide signature --

Everything happens for a reason. #1 reason: poor planning
WSSA #44

MMuddler Veteran Member • Posts: 4,526
Re: Buffer vs write speed a 30 second test

Stusteelhead wrote:

I think you are missing my point a bit, sports and wildlife definitely have similar but different needs that I understand. My point is a big buffer 25 plus shots and fast write speed allow you to shoot a lot of bursts over a short amount of time. no i don't just hold the trigger down for 30 seconds but what often happens is that a subject will fly in for only 30 seconds or less so the most shots I can get over that sort of time span is very important to me. The fact that the D300 is able to move that much data in 30 seconds means you rarely have to wait on the buffer, and therefore hopefully don't miss the shot. In sports you will often shoot longer bursts nearly filling the buffer over an action sequence, and need it to empty for the next, I get that, but speed is speed so I feel this test is important to both of us in that its representative of a cameras ability in moving data, like i said i would be curious to see how a D7000 does and the new D7100 with a fast card.

I really believe jfriend00 understands that how you shoot  is different than how a sports shooter might shoot, and that what you are doing is attempting to simulate your real work situation.

The point he is making is that your test procedure isn't optimal because 1) it won't be repeatable consistently since there will be variations in how long an individual holds the burst and how much inter-burst time he will employ, and 2) running a full 30 seconds (rather than cutting off as soon as the buffer fills) means you may be collecting/counting delayed shots that mean little on judging burst performance,  3) you are mixing measurement of buffer capacity with read-out performance, and 4) you are more vulnerable to differences between fps performance of different cameras (and battery) charge if the test is done by different people with  even the same model camera.

The take away from doing the test as a count of frames to fill the buffer... is that you know the buffer capacity. Furthermore, the time test to empty the full buffer let's you know how fast the write rate is .... and that lets you calculate the minimum time from "full" it takes to open up space for what is your typical burst.

-- hide signature --
OP Stusteelhead Senior Member • Posts: 1,135
Re: Buffer vs write speed a 30 second test

Makes sense and I guess my point is high fps and big buffer but slow write speed, still mean missed shots, I was just trying to show a lot of shots in 30 seconds (big buffer, high fps, and fast write speed),  for me means close to 100 shots and for you means quick clearing of the buffer. if the camera can only take 50 shots in  30 seconds that means it takes longer to clear the buffer also  And one day soon I imagine the write speed to the card will equal the write speed to the buffer therefore a small buffer may not be the issue it was in the past

-- hide signature --

--

www.raincoastphoto.com

Keyboard shortcuts:
FForum MMy threads