Notices


Reply
Thread Tools
Posts: 37 | Thanked: 67 times | Joined on Jun 2015 @ Munich, Germany
#1
So far it seems as all answers to the question "how good/fast is the SD slot?" are guesses. Some think the phone supports UHS, other don't and so on.

So I thought it's time to benchmark. Now I could provide a rpm but that's overkilled as all you need is the hdparm binary. Either download the sources from http://sourceforge.net/projects/hdparm/ and compile it for yourself (which is as simple as executing make) or use this precompiled one.

To benchmark with it do:
Code:
/path/to/your/hdparm -t --direct /dev/mmcblk1
Here's the first result:
Code:
# /home/nemo/horst-NFS/hdparm-9.48/hdparm -t --direct /dev/mmcblk1

/dev/mmcblk1:
 Timing O_DIRECT disk reads:  62 MB in  3.04 seconds =  20.40 MB/sec
The card in use is TS16GUSDU1.

Another benchmark could be to read from the card after dropping caches but first we would need a file to read... Ideas?

//EDIT: Just for reference here the speed of the internal storage:
Code:
# /home/nemo/horst-NFS/hdparm-9.48/hdparm -t --direct /dev/mmcblk0

/dev/mmcblk0:
 Timing O_DIRECT disk reads: 210 MB in  3.02 seconds =  69.57 MB/sec…
 

The Following 12 Users Say Thank You to V10lator For This Useful Post:
Posts: 415 | Thanked: 796 times | Joined on Jun 2011
#2
Code:
# /home/nemo/horst-NFS/hdparm-9.48/hdparm -t --direct /dev/mmcblk0

/dev/mmcblk0:
 Timing O_DIRECT disk reads: 210 MB in  3.02 seconds =  69.57 MB/sec…
Is that result from an N9 ? It should be that diferent from an N900 ? Mine shows only 18.24 MB/sec
Thanks for the code !
 
pichlo's Avatar
Posts: 4,933 | Thanked: 14,764 times | Joined on Sep 2012 @ UK
#3
Originally Posted by saponga View Post
Is that result from an N9 ?
What does the section name tell you The N9 does not have an SD card, BTW.

EDIT
FWIW, here are my results. First, from my Jolla:

Code:
[root@Dinghy hdparm-9.48]# /home/nemo/Dev/hdparm-9.48/hdparm -t --direct /dev/mmcblk0

/dev/mmcblk0:
 Timing O_DIRECT disk reads: 186 MB in  3.02 seconds =  61.66 MB/sec
[root@Dinghy hdparm-9.48]# /home/nemo/Dev/hdparm-9.48/hdparm -t --direct /dev/mmcblk1

/dev/mmcblk1:
 Timing O_DIRECT disk reads:  60 MB in  3.27 seconds =  18.38 MB/sec
[root@Dinghy hdparm-9.48]#
Next, for a comparison, from my N900:

Code:
Bianca:~# /home/user/MyDocs/Dev/hdparm-9.48/hdparm -t --direct /dev/mmcblk0

/dev/mmcblk0:
 Timing O_DIRECT disk reads:  70 MB in  3.09 seconds =  22.62 MB/sec
Bianca:~# /home/user/MyDocs/Dev/hdparm-9.48/hdparm -t --direct /dev/mmcblk1

/dev/mmcblk1:
 Timing O_DIRECT disk reads:  44 MB in  3.06 seconds =  14.39 MB/sec
Bianca:~#
Clearly I do not have the fastest SD cards in either of them. Not surprisingly, I went for the price, not quality

Last edited by pichlo; 2015-07-04 at 18:14.
 

The Following 9 Users Say Thank You to pichlo For This Useful Post:
Posts: 37 | Thanked: 67 times | Joined on Jun 2015 @ Munich, Germany
#4
Originally Posted by saponga View Post
Is that result from an N9 ? It should be that diferent from an N900 ?!
It's neither N9 nor N900 but Jolla.

Originally Posted by pichlo View Post
FWIW, here are my results.
They are worth a lot, at least for me, as this shows a slower card. Anyway, waiting for results of faster cards as I want to know if the slot can handle more than 20 MB/s.

I went for the price, not quality
I did that around 10 years ago, too, but then I had 10 out of 10 cards failing after a week (first 5, RMA, the new 5 failing again) from Kingston. Now it could be that the device using the card operated them slightly out of spec but the transcend replacements never failed (and even if they would have done I woul have had lifetime warranty - Other cards (can't remeber the brand) tended to fail when out of warranty). Since then I'm buying Transcend only. Not for speed (there are faster cards available from Transcend than the ones I'm using) but for quality.
 

The Following 5 Users Say Thank You to V10lator For This Useful Post:
seiichiro0185's Avatar
Posts: 247 | Thanked: 487 times | Joined on Nov 2007 @ Leipzig/Germany
#5
Here is the result from my Jolla:

Code:
[root@touko nemo]# ./hdparm -t --direct /dev/mmcblk1

/dev/mmcblk1:
Timing O_DIRECT disk reads:  58 MB in  3.08 seconds =  18.81 MB/sec
Card is a Sandisk 64GB Class 10 MicroSD
__________________
N800 -> N810 -> N900 -> N9 -> Jolla & TOHKBD -> Jolla C -> Xperia X F5122 (SFOS X)

http://www.seiichiro0185.org
 

The Following 5 Users Say Thank You to seiichiro0185 For This Useful Post:
jellyroll's Avatar
Posts: 405 | Thanked: 602 times | Joined on Apr 2012 @ Netherlands 020
#6
A result from my N900.
Attached Images
 
 

The Following 5 Users Say Thank You to jellyroll For This Useful Post:
Posts: 37 | Thanked: 67 times | Joined on Jun 2015 @ Munich, Germany
#7
@jeyroll Please note that there's a difference when using --direct. I try to get statistics from the card slot itself, so --direct is more close to metal.

//EDIT: For another test I just putted a sample from /dev/urandom of my desktop onto the sd card. Re-inserted into the phone and tried to read it:
Code:
$ dd if=random.sample of=/dev/null bs=4k
94033+1 records in
94033+1 records out
385162240 bytes (385 MB) copied, 19.3315 s, 19.9 MB/s
So again the 20 MB/s limit that we saw now with two class 10 cards. Has anyone any faster results?

BTW: When testing with such files like the one in my edit make sure (parts of) the file isn't in cache!

Last edited by V10lator; 2015-07-09 at 13:38.
 

The Following 2 Users Say Thank You to V10lator For This Useful Post:
Posts: 165 | Thanked: 350 times | Joined on Jul 2012 @ Germany - Potsdam
#8
The median from a range of 20.64 to 20.77 MB/s

Code:
hdparm -t --direct /dev/mmcblk1
/dev/mmcblk1:
 Timing O_DIRECT disk reads:  64 MB in  3.09 seconds =  20.73 MB/sec
SanDisk 128G Ultra microSDXC UHS-I Class 10 (max read: 48MB/s)
 

The Following 2 Users Say Thank You to meemorph For This Useful Post:
pichlo's Avatar
Posts: 4,933 | Thanked: 14,764 times | Joined on Sep 2012 @ UK
#9
Catching up with TMO after some absence and coming back to this thread. It seems that my el-cheapo 16 GB card is doing about as well as otber people's super-duper 64 or 128 GB. In other words, it is not my card that is the bottleneck. What else might it possibly be, I wonder

It might be interesting to compare the results for the same card between a Jolla and a desktop.
 

The Following User Says Thank You to pichlo For This Useful Post:
jellyroll's Avatar
Posts: 405 | Thanked: 602 times | Joined on Apr 2012 @ Netherlands 020
#10
Originally Posted by pichlo View Post
It seems that my el-cheapo 16 GB card is doing about as well as otber people's super-duper 64 or 128 GB..
Is 'otber' the word you really meant? Please don't do that in your rcS-late script.
 

The Following User Says Thank You to jellyroll For This Useful Post:
Reply

Thread Tools

 
Forum Jump


All times are GMT. The time now is 01:54.