EOS 7D CF Speed Test

[Lester Wareham]

As promised here is a speed comparison of the new 32Gb Sandisk card with two of the old ones.

Rather poor quality webcam video http://www.youtube.com/watch?v=Oc3MFy-mIis

For those that can't be bothered with videos the results are attached.

The 20 Mb/s Extreme III suffered from a reduced frame rate and number of main burst frames slightly and achieved a post buffer full, average frame rate of ~1 fps. The overall write rate was 17.4 Mb/s, 87% of its specified rate.

The 45 Mb/s Extreme IV bundled with the 7D managed the expected buffer number of frames and frame rate with a post buffer full, average frame rate of ~1.5 fps, the overall write rate was 39.7 Mb/s, similarly about 88% of the rated speed.

The new 60 Mb/s Extreme card again achieved the expected buffer depth and frame rate. In this case the post buffer full, average frame rate was a more useful 1.9 fps. The achieved write speed was 45 Mb/s, note this was 76% of the rated speed.

Of course the cards are of different sizes and the exact size of the RAW file may be better for one card's architecture than another. But is does seem that by the time the card rate gets to 60 Mb/s there is some onset of diminishing returns.

So although the 60Mb/s (400X) definitely shows an advantage over 45 Mb/s (300x) this result does leave me wondering if a 90 Mb/s (600x) card would be only slightly better than the 60 Mb/s.

I would be interested to see if someone with both 400x and 600x cards could repeat this sort of test.

[PacAce]

Very enlightening, Lester. Thanks for doing the test and sharing it with us. :)

Can you explain what "Burst Start" and "Burst End" mean, please, because it has me confused? If the shutter was held down for 10 seconds, I'd expect the burst end time less burst start time would be 10 seconds. Or does the burst end time indicate when the continuous firing of the shutter was first interrupted, i.e. when the buffers first filled up?

[Lester Wareham]

Hi Leo

This is the duration of the fully buffered main burst start and stop times worked out in an audio editor, so looking at the 60 Mb/s card this period was about two seconds at ~7.5 fps recording 15 frames, for the remaining 8 seconds the shutter was depressed a further 15 additional frames were captured in 12 groups with an average fps of 1.9 fps.

I attach the "sound waveform" overlay below, although I find this hard to interpret myself, but the pattern can discerned from the (admittedly poor quality) video sound track.

So looking at the bottom trace, this burst time (at maximum camera fps) is all the scurf in the first two seconds.

Of course it isn't easy to find the start point of the last frame in this burst, so no doubt there is a little measurement uncertainty present.

[nduralt]

Hi Lester, I recently purchased 4 of the 16 gig extreme 60mb/s cards and I really like them. I have used them for video in my 7d and shot bursts with it and have never really had any problems. Great cards in a good price range.

Its funny how your results tend to knock the card, but I've never had a real problem with them.

ND

[PacAce]

Hi Lester, I recently purchased 4 of the 16 gig extreme 60mb/s cards and I really like them. I have used them for video in my 7d and shot bursts with it and have never really had any problems. Great cards in a good price range.

Its funny how your results tend to knock the card, but I've never had a real problem with them.

ND

Lester's test is not knocking the 60 MB/s card. He does show that this card has a faster transfer speed than the other cards tested. But the transfer speed of the 60 MB/s card, according to his test, does seem to fall short of the advertised speed by a bigger margin compared to those of the other cards.

[Jon]

Lester's test is not knocking the 60 MB/s card. He does show that this card has a faster transfer speed than the other cards tested. But the transfer speed of the 60 MB/s card, according to his test, does seem to fall short of the advertised speed by a bigger margin compared to those of the other cards.Which suggests, to amplify on what Leo said, that the camera not the card, increasingly becomes the limiting factor as card speeds increase. This has been a consistent pattern throughout time, if you view Rob Galbraith's CF performance database (http://www.robgalbraith.com/bins/multi_page.asp?cid=6007). Remember that card speed ratings aren't measured in cameras; they're measured in standardized environments that exceed the capabilities of the cards being tested.

[nduralt]

Which suggests, to amplify on what Leo said, that the camera not the card, increasingly becomes the limiting factor as card speeds increase. This has been a consistent pattern throughout time, if you view Rob Galbraith's CF performance database (http://www.robgalbraith.com/bins/multi_page.asp?cid=6007). Remember that card speed ratings aren't measured in cameras; they're measured in standardized environments that exceed the capabilities of the cards being tested.

I see, good points PacAce and Jon. In the same way a computers bus speed may limit it's processing capability, or really how any choke point works.

Thanks for pointing out the Advertised aspect PacAce.

ND

[Lester Wareham]

Hi Lester, I recently purchased 4 of the 16 gig extreme 60mb/s cards and I really like them. I have used them for video in my 7d and shot bursts with it and have never really had any problems. Great cards in a good price range.

Its funny how your results tend to knock the card, but I've never had a real problem with them.

ND

Lester's test is not knocking the 60 MB/s card. He does show that this card has a faster transfer speed than the other cards tested. But the transfer speed of the 60 MB/s card, according to his test, does seem to fall short of the advertised speed by a bigger margin compared to those of the other cards.

Which suggests, to amplify on what Leo said, that the camera not the card, increasingly becomes the limiting factor as card speeds increase. This has been a consistent pattern throughout time, if you view Rob Galbraith's CF performance database (http://www.robgalbraith.com/bins/multi_page.asp?cid=6007). Remember that card speed ratings aren't measured in cameras; they're measured in standardized environments that exceed the capabilities of the cards being tested.

Thanks for looking everyone.

Yes my interpretation of the diminshing returns issue is the 7D is starting to have an effect.

An interesting point is although some reviewers state the 7D is designed to take advantage of UDMA 6 cards AFAIK canon only talk of UDMA and make no mention of modes or card rates.

So although a mode 6 90 Mb/s card may result in higher speeds it may not be terrific value for money, at the moment in the UK the mode 6 SanDisk cards are about twice the cost of the mode 5 (60 Mb/s).

[nduralt]

Lester, after watching your video a second time I can't help but ask if it was in raw/jpeg mode. My 7d usually gets hung up pretty bad when using both, and yours seemed to get hung up pretty bad there too after about 12 shots or so, I thought the buffer could hold more than that. Not that I am trying to imply anything at all.

Anyways... the test would have to be conducted with the 90mb/s card to prove this one, but I think it is possible that the Sandisk cards are at fault. To me it appears there is an appreciable increase in performance, so nothing can be said about why it is so marginal (ex. the camera has reached its operational limit). At this point I just have more questions than answers, and I personally don't want to drop a few hundred to test this theory... I guess I'll leave it to better people than myself. 2¢

[Lester Wareham]

Lester, after watching your video a second time I can't help but ask if it was in raw/jpeg mode. My 7d usually gets hung up pretty bad when using both, and yours seemed to get hung up pretty bad there too after about 12 shots or so, I thought the buffer could hold more than that. Not that I am trying to imply anything at all.

Anyways... the test would have to be conducted with the 90mb/s card to prove this one, but I think it is possible that the Sandisk cards are at fault. To me it appears there is an appreciable increase in performance, so nothing can be said about why it is so marginal (ex. the camera has reached its operational limit). At this point I just have more questions than answers, and I personally don't want to drop a few hundred to test this theory... I guess I'll leave it to better people than myself. 2¢

The camera was in raw only as per the result sheet.

I don't know how much you can read into the results, they are what they are.

If I repeated the test with a 8Gb 60mb/s card to match the size of the 45 mb/s one we may see the result in a better light, ie the diminishing returnes could be partly due to the different card architecture due to the factor of 4 difference in size.

But these are the cards I happen to have.

[hollis_f]

Interesting that the burst frame rate is so slow for the Extreme III. Was the lighting the same for each test? As we know the 7D's frame rate will drop in low light.

I found that the time to clear the buffer didn't drop as much as expected when going from 45 MB/s to 60 MB/s - backing up your suggestion that the 7D's write speed is limited to about 50 MB/s. Would love to hear from anybody who's tried something similar with a 90 MB/s card.

[Lester Wareham]

Interesting that the burst frame rate is so slow for the Extreme III. Was the lighting the same for each test? As we know the 7D's frame rate will drop in low light.

I found that the time to clear the buffer didn't drop as much as expected when going from 45 MB/s to 60 MB/s - backing up your suggestion that the 7D's write speed is limited to about 50 MB/s. Would love to hear from anybody who's tried something similar with a 90 MB/s card.

Yes the lighting was the same, you can probably see this from the video. As noted in the results sheet the camera was on manual 1/1000 f2.8 to keep the shutter speed up.

To give a reasonable amount of detail in the RAW (for a realistic file size) the ISO was ramped up to H and the lens focused on the books. AF operation is not on the shutter button so the was no AF delay. The file size (31-32 Mb) is perhaps a bit higher than a typical 7D RAW which I find are typically 25-27 Mb.

Yes the frame rate for the main burst did seem a bit lower for the Extreme III card, but the exact start and stop points are difficult to determine accurately.

More significantly the number of frames in this burst was 2 less, reasonable if the pipe emptying the camera buffer is being slowed by the card.

I am not making any claim about the 7Ds write speed, although this inference is certainly suggested by the results; we have to keep in mind all the cards had different sizes which may be significant.

I did play with some numbers which suggested a 90 Mb/s card might achieve 60-70 Mb/s, but I stress the word play. There is insufficient data and insufficient knowledge on my part of the insides of both card and camera for me to make even a ball park estimate.

What we need is someone to try this with a 32 Gb 90 Mb/s card.

[ticklemestephen]

thanks for the post

[dco31]

Thanks a lot for this test and all your comments.
I bought my eos 7d yesterday and now I can order cards.
Note that nowadays, in France, Extreme IV are more expensive than Extreme!

KRGDS

dco31

[Lester Wareham]

Thanks a lot for this test and all your comments.
I bought my eos 7d yesterday and now I can order cards.
Note that nowadays, in France, Extreme IV are more expensive than Extreme!

KRGDS

dco31

Enjoy your new 7D.

Yes the uk too has this odd price structure. I can not see any difference in the quality level between the new 60 and 90 mb/s cards and the extreem IV 45 mb/s, so it may just be down to larger numbers of units being sold amortizing the design NRE costs.

[RiaGurl]

As promised here is a speed comparison of the new 32Gb Sandisk card with two of the old ones.

Rather poor quality webcam video http://www.youtube.com/watch?v=Oc3MFy-mIis

For those that can't be bothered with videos the results are attached.

The 20 Mb/s Extreme III suffered from a reduced frame rate and number of main burst frames slightly and achieved a post buffer full, average frame rate of ~1 fps. The overall write rate was 17.4 Mb/s, 87% of its specified rate.

The 45 Mb/s Extreme IV bundled with the 7D managed the expected buffer number of frames and frame rate with a post buffer full, average frame rate of ~1.5 fps, the overall write rate was 39.7 Mb/s, similarly about 88% of the rated speed.

The new 60 Mb/s Extreme card again achieved the expected buffer depth and frame rate. In this case the post buffer full, average frame rate was a more useful 1.9 fps. The achieved write speed was 45 Mb/s, note this was 76% of the rated speed.

Of course the cards are of different sizes and the exact size of the RAW file may be better for one card's architecture than another. But is does seem that by the time the card rate gets to 60 Mb/s there is some onset of diminishing returns.

So although the 60Mb/s (400X) definitely shows an advantage over 45 Mb/s (300x) this result does leave me wondering if a 90 Mb/s (600x) card would be only slightly better than the 60 Mb/s.

I would be interested to see if someone with both 400x and 600x cards could repeat this sort of test.
would love to try this test, but how did you measure these figures? software? how do we do the same?

[Lester Wareham]

would love to try this test, but how did you measure these figures? software? how do we do the same?

Please do try the test. The more data points the better.

The methodology was very much a seen; set camera for RAW (I only shoot RAW), manual and a high shutter speed to keep the frame rate up (1/1000 here), set high speed shooting, set the iso to get a reasonable exposure and focus on some detail so the image complexity is representative and thus so is the file size. (ie don't shoot blank or oof frames)

Video the test from the back of the camera (I used the pc monitor web cam and mic array). Press the shutter for 10 seconds (timed off my wrist watch seconds hand), this fills the buffer and shows how the camera/card cope with the buffer full state.

Get your footage into a video editor. For monitoring the audio track you may find it best to export the audio to an audio editor that will display a sound waveform.

Determine the start of data write from the first shutter sound, the end of the buffer fill process from the last of the continuous shutter sounds. You can also use sound to work out the "post buffer full" shutter firing. (This is not that easy for the later sounds, but the first sound is easy, mS accuracy is not needed, correct to a few video frames will be good enough)

Use the video editor to note the time of the preview on the LCD and when the write light goes out.

Copy the files onto the pc, count the number of files and the total number of bytes (not the size on disk but the size of the file).

Then some arithmetic.

ie The card write speed is estimated from: <total number of Mb>/(<write light off time Seconds> - <first shutter sound start Seconds>)

Things like pre and post buffer full frames per second are obvious.

Lots of video and audio editors are usable, many free or relatively cheap. I used AV4YOU from Online Media Technologies http://avs4you.com/index.aspx.

Good Luck.

[nduralt]

would love to try this test, but how did you measure these figures? software? how do we do the same?

This is going to be very interesting indeed. Thanks for offering!