With the recent arrival of Canon’s new 5D Mark IV it was time to see how it was going to deal with my cards. Those of you who have photographed alongside me in the field (such as on our LuminOdyssey workshops) may have noticed that I pay particular attention to sorting out the technical side of my gear beforehand so I know what to expect in the field. One of these issues is which memory cards to put in my cameras.
The underlying speed of a card (in MB/s: megabytes/second) tells us a lot, and in general provides us with a metric to “rate” our cards (I generally arrange my cards so the fastest are used first, only going to slower/older cards if I run out of space during an outing). But in the field it’s a slightly abstract number. I have found there are a few numbers which are most-useful in terms of the shooting experience:
- Burst frame rate
The manufacturer advertises a number in terms of frames/second (fps), but the actual performance is not always the same.
- Seconds of buffer
How long we can shoot a burst for before the camera slows down (and the buffer count in the viewfinder reaches zero). Mind you, I try to not just “spray” a long burst of shots, but a series of shorter bursts can still fill up the buffer quickly.
Note that this is just derived from the frame rate and the number of shots in the buffer.
- The underlying card speed in frames/sec
This is really a measure of how fast the buffer empties after the available count has reached zero.
It should be obvious that these measures will be affected by the size of each photo, and with Canon’s losslessly-compressed CR2 files this can unfortunately vary quite a bit depending on the ISO and the complexity of the scene. But I find them useful measures nonetheless.
My own cards
When I buy a new memory card it’s usually one of the faster models available, and usually larger in capacity than the older cards. As a result my cards are a motley collection, with often only one example of each model. While it can be convenient to have huge fast cards and not need to change cards during an outing, the cost of this can be significant (especially when a new and improved card appears on the market soon after).
I’m interested in how fast each of my cameras will write to the cards. I also want to make sure the download speed in my card readers is as fast as possible in order to streamline my workflow, but the tests here are just about how fast my cameras write to the cards. To complicate matters, variations in the hardware and software of each camera model can result in some particular card models being at the head or the back of the pack in different cameras. So camera-specific testing has always been required.
Eventually what were once my best cards get retired, so my list of tested cards is not huge. Mind you there are a few cards in these results which are not my own cards: they were lent by friends curious about their own cards.
Starting with the EOS 5D Mark III most of my cameras have had a fast CompactFlash slot along with a slower SD slot. The 5D Mark III’s SD slot was limited to around 19 MB/s, but it was still useful to me. I have my cameras set to use one card at a time, but auto-switch to the other when required. This has allowed me to avoid the dreaded Card Full messages in the middle of an interesting action sequence. When the camera switches over to the slower card the camera’s buffer empties more slowly, and this is often what makes me notice that it’s switched. In the next break in the action I can then swap in a fresh faster card, switch back to that (and format it). Each time the fast card fills up, I’ll end up with some shots on the “overflow” card, but once they’re all imported into Lightroom they get sorted by capture time and it doesn’t matter. But as the SD cards and the cameras have improved, some of them are now faster than some of my older CompactFlash cards, so it helps to know which card to put in next (and whether to start treating the CF card as the “overflow”.
Note that if you instead shoot with the camera set to write the same files to both card slots, it’s the performance of the slower card which will define the overall performance of the camera, no matter how fast a CF card you have installed.
I shoot a long burst in high-speed continuous mode (usually on the order of 30 seconds) of RAW shots of a well-lit scene (in manual exposure with a shutter speed of at least 1/250s). Video recordings of the camera show the card activity light turning on and off (to a precision of around 1/30s) and looking at the audio track it’s easy to count the number and timing of the exposures. There’s usually some per-burst overhead in terms of how long the card activity light is on before the camera starts writing to the card, but this is easily estimated using a series of burst of different lengths, and then this overhead is subtracted from the timings. I try to do at least 3 seperate runs and average the results.
All lens correction features were turned off (I shoot in RAW and this is handled later). I did not test JPEG-only or RAW-plus-JPEG, as those have no interest for me in my own use of the cameras. I use my cameras with AI Servo AF almost all the time, although I use back-button focus so the AF is only actually engaged while the button is held down. In these tests AF was not active, although AI Servo was selected. This can sometimes affect the results (as will be seen), but matches the conditions in which I use these cameras.
Burst performance during actual shooting will depend on many factors. Shots can get delayed due to Anti-flicker timing if that’s engaged (you should leave it off except for those times you actually need it as it has other side-effects also) and they can get delayed depending on the AI Servo configuration (I have mine set to prefer having in-focus shots rather than just firing as fast as possible). But those factors should not have been at play in this testing in order to provide a baseline.
Times to write each set of images were accurate to within 1/30s, and we had at least 3 significant digits in the MB measurement. But even so the results varied a little. One of the factors in this would have been slight variations in the CR2 file size, but these were reasonably consistent and any outliers were discarded. The battery wasn’t returned to a precise charge level for each test, and many other such variables are presumably at play, and the precise results sometimes varied noticeably. As well as averaging the results of multiple tests for each card, I have rounded the results down to avoid false interpretation of precision in the table. But even a difference of several MB/s is probably going to be indiscernable to the photographer behind the camera.
EOS 5D Mark IV
As Canon’s latest 5D-series camera I hoped to see improvements in the card-writing speed. The camera is advertised as managing 7 fps, with 30 Mp stills, so those CR2 files are going to need a fast card. Indeed I did find the camera faster at saving to these cards than any previous 5D (or 7D) cameras.
If you’re reading these tables on a phone you might need to turn it sideways to see all the columns.
|Card||Burst fps||MB/s||Underlying RAW fps||Buffer seconds|
|Lexar 1066x 64GB CF||7||108||3.2||3.4|
|SanDisk Extreme Pro 160 MB/s 32GB CF||7||105||3.2||3.6|
|SanDisk Extreme Pro 160 MB/s 16GB CF||7||102||3.2||3.4|
|Lexar 1000x 32GB CF||7||101||3.2||3.7|
|Kingston Ultimate 600x 32GB CF||7||95||3.1||3.7|
|SanDisk Extreme Pro 90 MB/s 16GB CF||7||88||2.9||3.6|
|Lexar 1000x 16GB CF||7||84||2.8||3.4|
|Lexar 600x 16GB CF||7||83||2.7||3.4|
|Kingston Ultimate 600x 32GB CF||6*||83||2.7||4.7|
|SanDisk Extreme Pro 95 MB/s 64GB SD||7||77||2.5||3.1|
|SanDisk Extreme Pro 95 MB/s 32GB SD||7||77||2.5||3.4|
|SanDisk Extreme Pro 95 MB/s 16GB SD||7||77||2.5||3.4|
|Lexar 2000x UHS-II 64GB SD||7||76||2.3||4.0|
|SanDisk Extreme 90 MB/s 16GB SD||7||66||2.1||3.3|
|Lexar 2000x UHS-II 64GB SD||6*||65||2.0||2.9|
|SanDisk Extreme Plus 80 MB/s 16GB SD||7||65||2.1||3.3|
* You will notice two timings for each card in this camera. It turns out that when the battery (whether it’s an LP-E6N or an LP-E6) charge level drops below 50%, if you have AI Servo selected the frame rate drops from 7.0 fps to 6.0. This is hinted at on page 162 of the 5D Mark IV’s manual (although it mentions that the rate can drop to 5 fps with the LP-E6, I have only observed 6 fps with both battery types). A side-effect is that the buffer length sometimes increases significantly (both in number of frames and overall time). The underlying write speed of the card usually doesn’t change much, with a couple of exceptions such as the Kingston Ultimate 600x card.
As can be seen in the above table from the 5D Mark IV manual, the battery indicator shows when the level is below 50%, and if you have the battery indicator set to be visible in the viewfinder you’ll be able to monitor it without taking your eye from the viewfinder.
Do however note that it’s as yet unclear whether the lower battery level also has side-effects on things like AF tracking performance, so you might not want to keep all your batteries below 50% charge just to achieve a longer buffer. It’s reasonable to assume that Canon drops the frame rate for a reason (hopefully to maintain AF performance) even if they didn’t document it in the manual: keeping your batteries above 50% where possible should be a safe move.
Incidentally I did check: this drop in the burst rate is consistent even for RAW+JPEG or JPEG-only.
As can be seen, some of these cards achieve underlying write speeds faster than 3 fps, and with them you’ll find that if you’re shooting on continuous-slow (3 fps) you will effectively have an infinite buffer when shooting RAW (unless your CR2 files end up significantly larger than those in my testing, due to high ISOs and/or particularly detailed scenes). The per-burst overhead (the constant extra time the card activity light is on) in this camera seems to be 0.4 seconds.
EOS 5Ds R
Some people might not regard the 5Ds R as an “action” camera, but with 5 fps performance, huge 50 Mp files, and an AF system based on that of the 5D Mark III and 1Dx, it’s proven to be a workhorse camera in the field as well as in the studio. I’ve used this camera as my main body (sometimes in partnership with a 7D Mark II) for almost a year now, and it’s seen action in both polar regions, in rain, snow, and sun. With the huge files produced by this camera, selecting the fastest possible cards certainly makes a huge difference in how fast the buffer empties.
In terms of MB/s, the card writing speed in this camera is similar to the EOS 7D Mark II (which I no longer have so won’t be posting numbers for it). But with the much larger files the buffer obviously empties more slowly. The per-burst overhead (extra time the card activity light is on) in this camera seems to be 1.25 seconds. The burst frame rate for the 5Ds R was consistently 5.0 fps across all these cards, so this table is shorter than the 5D Mark IV’s.
|Card||MB/s||Underlying RAW fps||Buffer seconds|
|SanDisk Extreme Pro 160 MB/s 16GB CF||93||1.7||3.6|
|Lexar 1000x 32GB CF||91||1.7||3.6|
|SanDisk Extreme Pro 160 MB/s 32GB CF||88||1.7||3.4|
|Lexar 1066x 64GB CF||87||1.6||3.4|
|Kingston Ultimate 600x 32GB CF||84||1.6||3.6|
|SanDisk Extreme Pro 90 MB/s 32GB CF||84||1.6||3.2|
|Lexar 1000x 16GB CF||77||1.4||3.4|
|SanDisk Extreme Pro 95 MB/s 64GB SD||76||1.4||3.2|
|SanDisk Extreme Pro 95 MB/s 32GB SD||75||1.4||3.0|
|SanDisk Extreme Pro 95 MB/s 16GB SD||70||1.3||3.0|
|Lexar 2000x UHS-II 64GB SD||67||1.2||3.0|
|SanDisk Extreme Plus 80 MB/s 16GB SD||63||1.1||3.0|
|Lexar 600x 16GB CF||50||0.9||3.0|
Comments and conclusions
It was very interesting to see that a few cards are dramatically faster in the 5D Mark IV than in previous cameras. Particular examples are the SanDisk Extreme Pro “90 MB/s” and Lexar 600x cards. At the same time the Lexar 1066x 64GB and SanDisk “160 MB/s” 32GB cards were slower than expected in the 5DsR. But such uneven differences in card performance across different cameras is not unusual at all, which is why these tests are important.
I’m glad I now have this information available to help me choose the best memory cards (and batteries) to place in the cameras in the field. Hopefully this information is also of use to other 5D Mark IV and 5Ds R shooters.