How fast can jolla phone write to / read from sd cards

asked 2015-01-06 15:51:30 +0300

ralooyar gravatar image

updated 2015-01-06 15:52:24 +0300

Reading the Jolla SD-Card Wiki, it is not really clear, whether jolla phone supports UHS-I speed or not. There are people mentioning that UHS-I microSD cards work, but those cards might just fall back to non-UHS-speed.

So it would be very interesting to see some real-live speed measure results: How fast can your jolla phone write to and read from your sd-card?

Of course, this will vary, based on the SD-card you use. But if you have a real fast SD-card, read/write performance will be limited by the jolla phone. And this is the value, which I am interested in: How fast can the jolla phone write to / read from a SD-card.

So people can buy SD-cards according to this performance values.

Any ideas for good and simple to implement speed measurement?

I thought of testing the writing speed with:

dd if=/dev/urandom of=/home/nemo/speed-test bs=1M count=100

But '/dev/urandom' seems to be to slow. It gives me writing speeds of 2,6 MB/s (40 seconds for 100 MB).

Using '/dev/zero', I fear, this give problems due to easy compressible content. In fact:

dd if=/dev/zero of=/home/nemo/speed-test bs=1M count=100

gives "138 MB/s" for writing on the internal jolla flash. And 66,2 MB/s for reading from the internal jolla flash via

dd if=/home/nemo/speed-test of=/dev/null bs=1M count=100

This look very strange to me.

Nevertheless one should clear the cache in between via:

devel-su
echo 3 > /proc/sys/vm/drop_caches

or you will get reading speeds around 602 MB/s. ;)

edit retag flag offensive close delete

Comments

1

Why does 66/138 MB/s look strange to you? Also, compressible data is just an issue if the controller would compress (Sandforce controllers do, some others do, some don't). I don't know if the Jolla has a compressing controller?

I think the problem there may be related to the size of the test. 100 MB will most likely be buffered to a good extent on the OS level, you should run at least a gigabyte of data to eliminate that artifact.

sidv ( 2015-01-06 16:08:23 +0300 )edit

@sidv: 66/138 MB/s looks strange to me, becaus I would expect the writing speed (138 MB/s) would be slower than reading (66 MB/s). But you are right. 100 MB seems to be a little small.

ralooyar ( 2015-01-11 18:49:09 +0300 )edit