I have been having a close look at some SD cards I had on my desk and tried to test them. Please note that this was done just to try to understand the differences between cards that carry similar specifications, not to take position on the quality of the DR-70D write-to-card routines.
To understand the test results, one must have a reference : recording at 24/96 4 channels, with 20% overhead to manage the file system should require something like 1.5MB/s of bandwidth.
Tests were run using the gnome_disks_utility (disks) on a Fedora 22 linux machine. This tests the random read and write speed of the cards. It was set to 250 samples of 10 MB each.
The 10 cards tested are in fact 2 each of 5 different models :
- 2 SanDisk 16GB SDSDXPA-016G-X46 on the Tascam DR-70D recommended list
- 2 SanDisk 64GB SDSDXPA-064G-X46 on that same recommended list
- 2 Transcend 32GB of 2013 no speed marked
- 2 Transcend 32GB TS32GSDHC10E of 2014 C10, 200x
- 2 Transcend 32GB TS32GSDHC10U1E of 2015 C10, 600x
All cards bought from Amazon (and sold by Amazon, no marketplace sellers).
Here are the results (red line is the write rate in MB/s, blue line is the read rate in MB/s, the green scatter plot is the access time in msec) :
SanDisk 16GB, very lightly used (just some tests) :
SanDisk 16GB, new out of the box :
SanDisk 64GB, new out of the box :
This card is reported as 32GB by the program, it has been checked with f3 and is indeed a 64GB card with all sectors working correctly.
another SanDisk 64GB, new out of the box :
Same tests were performed with f3, the card is really a 64GB card.
All these cards show a very regular pattern that does not change during the test and a minimum write speed that is much higher than the 1.5MB/s that is needed.
Now lets look at some Transcend cards that have served me well in photo applications :
Transcend 32GB, used but not abused, no speed indication on the card (bought in 2013) :
another Transcend 32GB, used but not abused, no speed indication on the card (bought in 2013) :
Let's get a little more current with Class 10 cards :
Transcend 32GB, used but not abused, Class 10, 200x (bought in 2014) :
another Transcend 32GB, used but not abused, Class 10, 200x (bought in 2014) :
And even more current, Transcend Class 10 600x UHS-I cards :
Transcend 32GB, used but not abused, Class 10, 600x, UHS-I (bought in 2015) :
another Transcend 32GB, used but not abused, Class 10, 600x, UHS-I (bought in 2015) :
With these cards, performance is progressively reduced during the test (obviously, there is buffering at play at the beginning). Write rates may be insufficient at times to ensure a 1.5MB/s bandwith. Apart from the first type (the older one) that seems to guarantee a 6MB/s bandwith, they do not seem to be a good choice for a multitrack recorder. It is strange that the more recent cards that are sold as better (with more badges such as Class 10, UHS-I, ...) have in fact much worse write speed performance. Access time can be very high on rare occasions at 6 to 8 msec.