Reading mode: Light Dark

What the... Memory Card Speed Test with 1DsMkIII --> SURPRISE!

Go to last post
Earthlight

Earthlight

Senior Member
Messages
3,249
Reaction score
97
Location
FI
Introduction

So, I thought I'd carry out a little memory card speed test with my new 1DsMkIII.

Today I bought two new cards in hopes of getting as fast buffer flush as possible and maybe even longer bursts. Then I proceeded to compare my new acquisitions against my older cards.

Keep reading as THE RESULTS PROVED SHOCKING AND TOTALLY UNEXPECTED...

Here are the contestants (sorry about the crappy iPad photo):



My new cards are on the left, the 400X Transcend 32GB CF and below it the SanDisk Extreme 32GB SD. The other cards I just scoured from my camera bag.

The testing procedure
  • Canon EOS 1DsMkIII
  • Manual exposure 1/500s at full aperture
  • High speed continuous drive
  • Manual focus
  • Format each card before the test burst
  • Quality: RAW (no jpeg recorded)
  • Wifey clocking the elapsed time with my iPad and the Clock Pro HD app.
  • Timing begins as I say "one, two, three, NOW" and ends as the red write indicator LED goes dark.
  • Shutter button lifted as soon as the continuous burst halted
The results

All the cards from the cheapest to the most expensive managed an exactly 13 frame burst until slowing down (continuous burst suddenly halted). No more no less. The fastest times are in bold.

Compact flash

Transcend 400X 32Gb UDMA7 --> 13 frames, 20,3 seconds
SanDisk Ultra 30MB/s 16Gb --> 13 frames, 20,3 seconds
SanDisk Ultra II 15MB/s 8Gb --> 13 frames, 19,2 seconds

Secure Digital

SanDisk Extreme 45MB/s SDHC class 10 UHS-I 32Gb --> 13 frames, 22,8 seconds
Transcend SDHC class 10 32GB --> 13 frames, 20,3 seconds
Kingston SDHC class 4 16 Gb --> 13 frames, 36,3 seconds

And out of general interest two cards together:

Transcend 400X 32Gb UDMA7 and
SanDisk Extreme 45MB/s SDHC class 10 UHS-I 32G

--> 13 frames, 23,7 seconds

Conclusion

I do not know what to make of this. Out of the CF cards the fastest one was the oldest and nominally slowest . Granted, the capacities were different but certainly a bigger newer generation card with way faster specs should not be slower than a smaller old tech card?

In the SD competition I find it unexpected that the cheaper 32GB Transcend beats the SanDisk Extreme .

About the only thing that makes sense to me here is the painfully slow 36 seconds for the Kingston SDHC class 4 16 Gb card.

What are we seeing here? Are smaller old tech cards usually faster than newer bigger capacity cards? Is the DsMkIII unable to take advantage of the new tech, perhaps getting even slower buffer flush because of it? Caveat emptor?

INVITATION

Do you own a DsMkIII and some 32GB cards such as the Sandisk Extreme or Extreme Pro CF? Or some other type of a 32GB CF or SD?

Please, repeat the prosedure with those cards and report your findings here. This should be interesting.

Pre-emptive strike against the inevitable trollsy accusations :-D
  • Yes I do realise this is not the biggest issue in the world
  • No, I do not lose sleep over this
  • No, I am not steaming but enjoying myself, this is interesting
  • Yes, I have better things to do in life but had some spare time today and came up with this idea, hehe.
Earthlight

--

http://jari.pic.fi
 
Here are the Dpreview results:

http://www.dpreview.com/reviews/canoneos1dsmarkiii/13

See, I did not come even close their 11 sec flush times with my current cards. And they tested it in August 2008 with a SanDisk Ducati card.

And they even got a whopping 18 image burst with the then fastest card. I could not get over 13 frames even once.

I wonder what gives. This is really interesting.

Earthlight

--

http://jari.pic.fi
 
I just picked up a 1D4 and am looking for buffer flush speed/card performance as well. One item I found pointed to poor performance with UDMA 7 cards at a back level of the firmware.

That same web item showed a 6 second flush with the newer 1000x cards, something I'm considering.
 
Thanks, who knows. I tried at ISO100 with a lens cap on and managed to increase the burst to 15 frames while at the same time shaving about two seconds off the flush time.

Of course, ISO100 files have less noise and dark frames like that compress better.

Still not even close to the Dpreview measurements. Quite puzzling.

Earthlight

--

http://jari.pic.fi
 
If the image was totally black then this is expected.
 
ossme said:
If the image was totally black then this is expected.
Click to expand...
The shots & times reported in the OP were at ISO400 and I shot the kitchen table in the afternoon light. Deleted the files already.

ISO100 and the lens cap on increased the burst length to 15 frames while shaving about two seconds off.

Should I shoot white frames in bright light to max the numbers? I wonder what Dpreview shot in their test. (Not that I bought the camera to do that, just genuinely interested).

Earthlight

--

http://jari.pic.fi
 
Have you tried to turn the noise reduction off ?
 
I always heard it is not the speed of the card. when clearing the buffer and burst shooting. It is a bottle neck of the write speed of the camera?

And even the low speed cards are pretty fast. The difference is writing to the computer. Anyway I use a cheap card and get just as good as my buddies on burst shooting.
 
There is no 'shocking news' at all. See? I told you so in the other post, burst length has nothing to do with cards and everything to do with camera RAM size and settings. You didn't listen and wasted money on fast cards.

No matter how fast your card is, it will not keep up with continuous shooting speed. When camera RAM buffer is full it will jam.

The more reasonable thing to do is learn how the buffer system works and what settings affects the buffer length and adjust to field conditions accordingly.

On the right side of your VF, below meter, there is a number indicating remaining buffer size, use it to your knowledge so you won't get 'surprises'
 
BozillaNZ said:
No matter how fast your card is, it will not keep up with continuous shooting speed. When camera RAM buffer is full it will jam.

The more reasonable thing to do is learn how the buffer system works and what settings affects the buffer length and adjust to field conditions accordingly.
Click to expand...
This makes a great deal of sense. Can you point us to a place where we might get more info on things which affect buffer size and writing time?

I, too, was ready to go out and pay big bucks to get "fast" cards.

Jerry
 
Are we talking about burst until full or clearing the buffer?

The number of frames until the camera buffer is full will not change regardless of CF card you use. The buffer memory is much faster than any memory available today.

The amount of time it takes the camera to clear the buffer will change depending on how fast the card is. There is a limit to this which is how fast the camera can transfer the images from the buffer to the card.
 
What firmware you are currently running ?
 
garyknrd said:
I always heard it is not the speed of the card. when clearing the buffer and burst shooting. It is a bottle neck of the write speed of the camera?
Click to expand...
It depends on the camera - whether it supports fast cards or not. So it depends on both. My friend gets a totally different performance out of his 7D with fast and slow cards.

But faster cards should not be slower than slow cards. That is the surprise here.
garyknrd said:
And even the low speed cards are pretty fast. The difference is writing to the computer. Anyway I use a cheap card and get just as good as my buddies on burst shooting.
Click to expand...
It depends. Do you fill the buffer? The flush time will depend on the card performance if the camera supports sufficiently fast cards.

Earthlight

--

http://jari.pic.fi
 
BozillaNZ said:
There is no 'shocking news' at all. See? I told you so in the other post, burst length has nothing to do with cards and everything to do with camera RAM size and settings. You didn't listen and wasted money on fast cards.
Click to expand...
No. You misunderstand. See the Dpreview test that I linked above if you do not believe me.

The camera buffer is of course a matter of fixed hardware but if the card can receive data rapidly already during, say, a three second burst, then in effect you get more consecutive frames.

This is not the "shocking news" (how's that for reader motivation in the beginning of a long post) I referred to. I meant the fact that the nominally slower cards trumped the faster cards. Surely you too find thid a little strange?
BozillaNZ said:
No matter how fast your card is, it will not keep up with continuous shooting speed. When camera RAM buffer is full it will jam.
Click to expand...
See above. What is more, my friend gets significantly better performance from fast cards on his 7D. It is not fiction or marketing BS. We just tested it. More frames and faster flush times. And Dpreview confirmes this in their very own 1DsMkIII test. Check it out.
BozillaNZ said:
The more reasonable thing to do is learn how the buffer system works and what settings affects the buffer length and adjust to field conditions accordingly.
Click to expand...
Practical real life shooting necessitates this. That is a given. It is obvious. But better performance helps in real life too.
BozillaNZ said:
On the right side of your VF, below meter, there is a number indicating remaining buffer size, use it to your knowledge so you won't get 'surprises'
Click to expand...
That is not the surprise I was referring to. That much should be clear. When shootin I always observe this in case the situation warrants that.

Anyway, the DsMkIII is so much faster than the MkII that I longer will fill the buffer in 99,9% of my shooting scenarios. I have already benefited from the switch.

But since Dpreview got vastly better measuremenst already in 2008 with the then fast cards, do you not find it even slightly curious that I cannot replicate the results with the fast cards of 2012?

I'm done legitimating this any further.

Earthlight

--

http://jari.pic.fi
 
Thanks, I had a quick look at that yesterday but could not find any info pertaining to the puzzle at hand. Did you? I'd appreciate a reference to a specific page or paragraph.

Thanks again!

Earthlight

--

http://jari.pic.fi
 
If you test things properly, you should first understand what the word 'significant' means in science. It means that small differences could be the other way round if you test again, so they should not be considered real differences. I think the differences you found are probably not significant from a scientific perspective. How many times did you test? One test means nothing. Your test method is also questionable. Did you only count frames/second? That is unreliable because the file sizes may also vary a little. Did you consider to measure MB/sec, which is more reliable because that does take those small file size differences into account?

Your results do not surprise me at all. The 1Ds III does not support UDMA, so it should not be significantly faster with UDMA cards. That's exactly what you found. In fact, the UDMA controller in the cards might even slow it down just a little bit, but that's pure speculation on my part.

--
Johan
http://www.johanfoto.com
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top