SD Card Benchmarks


198 posts   Page 6 of 8   1 ... 3, 4, 5, 6, 7, 8
by rdatoc » Thu Sep 06, 2012 1:10 pm
stevepdp wrote:
rdatoc wrote:Posted a reply with my SD and SSD benchmarks but it seems to have fallen into a black hole...


All new accounts must have their first post approved by a mod.

Yes, I realized that when I finally noticed the message after making my second post. Hard to stop gazing at the flashing LEDs on the Pi sitting next to the PC. :D
Posts: 4
Joined: Wed Sep 05, 2012 10:21 pm
by BenPope » Tue Sep 11, 2012 3:09 pm
My iozone3 scores look like this:

Code: Select all
                                                            random  random
              KB  reclen   write rewrite    read    reread    read   write
           51200       4    1442    1733     5617     5626    5548     783
           51200     512    6810    7519    21921    21923   21898    2118
           51200   16384    8114    8826    22408    22407   22405   10626

8GB Sandisk Ultra 30MB/s Class 6
Posts: 4
Joined: Sun May 20, 2012 8:33 am
by Rain_1 » Tue Sep 18, 2012 2:29 pm
Hey there everyone! I just registered an account for this message.

Anyway, I recently received my Raspberry Pi and, as usual, I made a purchase without doing the proper research: I bought a "Kingston 32 gig Class 10 ultimate X 100x" card for my Pi (Pictured here: http://i.imgur.com/JFYmN.jpg) and, I gotta say, the benchmark on it is kinda terrible. Here are the header information and iozone report:

Code: Select all
==> /sys/block/mmcblk0/device/cid <==
413432534433324730008006a800c600

==> /sys/block/mmcblk0/device/csd <==
400e00325b590000edcf7f800a400000

==> /sys/block/mmcblk0/device/scr <==
0235800000000000

==> /sys/block/mmcblk0/device/fwrev <==
0x0

==> /sys/block/mmcblk0/device/hwrev <==
0x3

==> /sys/block/mmcblk0/device/manfid <==
0x000041

==> /sys/block/mmcblk0/device/oemid <==
0x3432

==> /sys/block/mmcblk0/device/name <==
SD32G

==> /sys/block/mmcblk0/device/serial <==
0x008006a8
Code: Select all
                                                            random  random    bkwd   record   stride
              KB  reclen   write rewrite    read    reread    read   write    read  rewrite     read   fwrite

frewrite   fread  freread
           51200       4     368     380     3706     3709    3565     178
           51200     512    4107    5876    21092    21102   21077     533
           51200   16384    4843    8021    21586    21655   21626    7172
I'm mostly using my Pi to run Raspbmc, streaming videos off my main computer. Video playing is working properly, but navigation is slow. I'm now wondering if these atrocious write speeds on the 4k block are the main culprit? I don't think it makes much sense, as I don't "write" on the card when I'm browsing, do I? I get 12MB/sec reading and 10MB/sec writing, according to hdparm (read) and dd (write).

I also tested a Kingston 2gig MicroSD that I had laying around. I don't even know the actual calssficiation of this card, but according to hdparm and dd, I'm getting 14MB/sec read and 5.6MB/sec write speeds, so I'm guessing it's a Class 6 card.

Here's the header information and iozone report for that random 2gb card:
Code: Select all
==> /sys/block/mmcblk0/device/cid <==
02544d534430324741a7766149009400

==> /sys/block/mmcblk0/device/csd <==
002e00325b5a83a9ffffff8016800000

==> /sys/block/mmcblk0/device/scr <==
02a5000020021102

==> /sys/block/mmcblk0/device/fwrev <==
0x1

==> /sys/block/mmcblk0/device/hwrev <==
0x4

==> /sys/block/mmcblk0/device/manfid <==
0x000002

==> /sys/block/mmcblk0/device/oemid <==
0x544d

==> /sys/block/mmcblk0/device/name <==
SD02G

==> /sys/block/mmcblk0/device/serial <==
0xa7766149
Code: Select all
                                                            random   random
              KB  reclen   write rewrite    read    reread    read   write   
           51200       4     893    1051     4585     4602    4482       8
           51200     512    5411    5388    21904    21871   21885    1102
           51200   16384    5471    5488    22244    22276   22217    5768


As you can see, the Random Write is pretty baffling. I didn't play much with it, but I noticed this random 2gig card is overall slower than my 32gig one.

With this build of iozone and list of commands, it's pretty easy to determine if a card is good or not. A database on these really has to come up. :D

Commands used:
Code: Select all
Sequential read:
sudo hdparm -t /dev/mmcblk0p3

Sequential Write:
dd count=1k bs=1M if=/dev/zero of=test.img

iozone settings:
iozone -e -I -a -s 50M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2
 
Card Header:
head /sys/block/mmcblk0/device/{cid,csd,scr,*rev,manfid,oemid,name,serial}
Posts: 2
Joined: Tue Sep 18, 2012 6:12 am
by gabrielmagno » Sat Sep 29, 2012 1:27 am
Card: SanDisk microSDHC 16 GB Class 4 with Adapter
Model: SDSDQM-016G-B35A

Boot time:
Code: Select all
2012-09-18-wheezy-raspbian.img:
- 25 seconds TTY without USB devices
- 35 seconds LXDE with USB Mouse and Keyboard


CrystalDiskMark Benchmark:
Code: Select all
-----------------------------------------------------------------------
CrystalDiskMark 3.0.1 (C) 2007-2010 hiyohiyo
                           Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

           Sequential Read :    19.948 MB/s
          Sequential Write :     5.828 MB/s
         Random Read 512KB :    19.606 MB/s
        Random Write 512KB :     1.857 MB/s
    Random Read 4KB (QD=1) :     3.508 MB/s [   856.4 IOPS]
   Random Write 4KB (QD=1) :     1.696 MB/s [   414.0 IOPS]
   Random Read 4KB (QD=32) :     3.608 MB/s [   880.8 IOPS]
  Random Write 4KB (QD=32) :     1.360 MB/s [   331.9 IOPS]

  Test : 50 MB [G: 0.6% (0.1/14.8 GB)] (x5)
  Date : 2012/09/28 14:14:53
    OS : Windows 7 Home Premium Edition SP1 [6.1 Build 7601] (x86)
 
Posts: 3
Joined: Thu May 24, 2012 4:08 pm
by DooMMasteR » Mon Oct 15, 2012 2:14 pm
Card: SanDisk MicroSDHC 8GB Class4
Code: Select all
dd-write: 5,3 MB/s
dd-read: 9,2 MB/s

Code: Select all
   KB  reclen   write rewrite    read    reread    read   write    read
51200       4    5585    5119   910124   967806  803528    5692
51200     512    5518    5542   696968   751205  748416    5648
51200   16384    5437    5432   731110   691433  678079    5673
Posts: 1
Joined: Mon Oct 15, 2012 2:00 pm
by Andy Woolford » Tue Oct 30, 2012 10:01 am
This is a SanDisk SDHC 32GB Class 4 device
Sn: BM1215622027G (printed on back)

-----------------------------------------------------------------------
CrystalDiskMark 3.0.2 x64 (C) 2007-2012 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 10.169 MB/s
Sequential Write : 6.682 MB/s
Random Read 512KB : 10.012 MB/s
Random Write 512KB : 1.427 MB/s
Random Read 4KB (QD=1) : 2.117 MB/s [ 516.9 IOPS]
Random Write 4KB (QD=1) : 0.654 MB/s [ 159.8 IOPS]
Random Read 4KB (QD=32) : 1.968 MB/s [ 480.4 IOPS]
Random Write 4KB (QD=32) : 0.656 MB/s [ 160.2 IOPS]

Test : 50 MB [I: 31.1% (23.2/74.4 MB)] (x5)
Date : 2012/10/30 9:55:37
OS : Windows 7 Professional SP1 [6.1 Build 7601] (x64)
Posts: 12
Joined: Fri Jun 22, 2012 1:26 am
by mkamelg » Wed Nov 07, 2012 5:45 pm
One card from new (2012) family called EXCERIA, from Japanese brand Toshiba.

Family: http://www.toshiba-memory.com/en/exceria_sd_cards.html (overview and technical data)

Tested on: Sony VAIO VPC-F23S1E with build-in reader (Ricoh PCIe SDXC/MMC Host Controller). SD card taken straight from the box, filesystem on card is FAT32.

Card model: Toshiba EXCERIA Type HD 16 GB SDHC Class 10 UHS-I

Benchmark:

Code: Select all
-----------------------------------------------------------------------
CrystalDiskMark 3.0.2 x64 (C) 2007-2012 hiyohiyo
                           Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

           Sequential Read :    43.978 MB/s
          Sequential Write :    29.650 MB/s
         Random Read 512KB :    43.509 MB/s
        Random Write 512KB :     5.469 MB/s
    Random Read 4KB (QD=1) :    15.137 MB/s [  3695.6 IOPS]
   Random Write 4KB (QD=1) :     0.431 MB/s [   105.2 IOPS]
   Random Read 4KB (QD=32) :    17.121 MB/s [  4179.9 IOPS]
  Random Write 4KB (QD=32) :     0.448 MB/s [   109.3 IOPS]

  Test : 50 MB [E: 0.0% (0.0/14.7 GB)] (x5)
  Date : 2012/11/07 16:46:00
    OS : Windows 7 Home Premium Edition SP1 [6.1 Build 7601] (x64)


Additional photos:

Front of box - http://www.pixbang.com/images/img0140.jpg
Back of box - http://www.pixbang.com/images/img0141.jpg
Front of card - http://www.pixbang.com/images/img0144.jpg
Back of card - http://www.pixbang.com/images/img0145.jpg
Posts: 1
Joined: Wed Nov 07, 2012 4:15 pm
by teckel » Sun Nov 18, 2012 5:36 am
SanDisk Extreme 16 GB SDHC Class 10 UHS-1 45MB/s SDSDX-016G-X46
$17 including delivery at Amazon.

Code: Select all
              KB  reclen   write rewrite    read    reread    read   write
           51200       4    1515    1592     4649     4639    4337     648
           51200     512   20270   18558    21742    21788   21684    4166
           51200   16384   16883   21262    22269    22230   22225   21185


These seem to be some of the best numbers reported for a SD card. I believe the final two "random" columns are the most telling of real-world speed. System boot time including keyboard, mouse, and wifi is 25.889967 seconds (512 KB version, medium 900MHz overclocking).

Tim
Posts: 3
Joined: Sun Nov 18, 2012 5:12 am
by teckel » Tue Nov 20, 2012 3:25 pm
What's the point of showing crystalmark benchmarks in Windows? We should only be showing benchmark numbers running on the RasPi, as they can be totally different than on Windows. A card may look quite good on Windows, but be much slower or not work at all on the RasPi.

Tim
Posts: 3
Joined: Sun Nov 18, 2012 5:12 am
by Drauger » Wed Dec 05, 2012 1:35 pm
Posts: 1
Joined: Wed Dec 05, 2012 1:30 pm
Location: Russia
by toysareforboys » Mon Dec 17, 2012 1:20 am
Bringing this thread back from the dead :)

I, like many others, bought a Kingston SD card (UltimateX 100x, class 10, 32gb), and OMG, the 4k random write performance :( totally garbage.

So I went out and bought a Patriot LX Class 10 card 32gb, almost equally garbage!

So then I finally bought Patriot EP Class 10 card, 32gb. OMG Pi performance is WAY better (it's a web server/php using Pancake and webmail/smtp with Citadel). My Pi status page: http://toysareforboys.no-ip.info:8000/

Image
LICK FOR HIGH RES!


So, I downloaded the .deb file linked earlier in this thread (apt-get iozone3 = not found), and ran the command: iozone -e -a -s 50M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2

But I'm getting completely different results than other people posted here? Almost like mine is in bytes instead of k? lol.
Code: Select all
pi@raspberrypi ~ $ iozone -e -a -s 50M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2
        Iozone: Performance Test of File I/O
                Version $Revision: 3.397 $
                Compiled for 32 bit mode.
                Build: linux

        Contributors:William Norcott, Don Capps, Isom Crawford, Kirby Collins
                     Al Slater, Scott Rhine, Mike Wisner, Ken Goss
                     Steve Landherr, Brad Smith, Mark Kelly, Dr. Alain CYR,
                     Randy Dunlap, Mark Montague, Dan Million, Gavin Brebner,
                     Jean-Marc Zucconi, Jeff Blomberg, Benny Halevy, Dave Boone,
                     Erik Habbinga, Kris Strecker, Walter Wong, Joshua Root,
                     Fabrice Bacchella, Zhenghua Xue, Qin Li, Darren Sawyer.
                     Ben England.

        Run began: Sun Dec 16 19:59:34 2012

        Include fsync in write timing
        Auto Mode
        File size set to 51200 KB
        Record Size 4 KB
        Record Size 512 KB
        Record Size 16384 KB
        Command line used: iozone -e -a -s 50M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2
        Output is in Kbytes/sec
        Time Resolution = 0.000001 seconds.
        Processor cache size set to 1024 Kbytes.
        Processor cache line size set to 32 bytes.
        File stride size set to 17 * record size.
                                                            random  random    bkwd   record   stride
              KB  reclen   write rewrite    read    reread    read   write    read  rewrite     read   fwrite frewrite   fread  freread
           51200       4   12260   13074   162062   178890  246140   12219
           51200     512   11958   14390   182284   204710  249234   11851
           51200   16384   14003   16545   151540   159875  159931   16120

iozone test complete.
Any ideas? :(

-Jamie M.
Seagate GoFlex Home, 1.2GHz ARM (kirkwood), 128MB RAM, Gigabit Ethernet, SATA2. Sandisk Extreme 120GB SSD running Arch ARM Linux 3.6.11-0. nginx + php-fpm = LIVE STATUS hosted right on the SGFH!! http://tafb.yi.org
User avatar
Posts: 136
Joined: Thu Dec 06, 2012 11:01 pm
by toysareforboys » Mon Dec 17, 2012 5:12 am
Ahhh. I was reading the manual for iozone and it said to make sure, for accurate results, to make the -s switch size be TWICE your ram size. My Pi has 512mb so figured minus the 16mb gpu share I'd make it 1000M (for 1gb test file). Worked way better:

Code: Select all
 pi@raspberrypi ~ $iozone -e -a -s 1000M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2 -Rb chart.wks
        Iozone: Performance Test of File I/O
                Version $Revision: 3.397 $
                Compiled for 32 bit mode.
                Build: linux

        Contributors:William Norcott, Don Capps, Isom Crawford, Kirby Collins
                     Al Slater, Scott Rhine, Mike Wisner, Ken Goss
                     Steve Landherr, Brad Smith, Mark Kelly, Dr. Alain CYR,
                     Randy Dunlap, Mark Montague, Dan Million, Gavin Brebner,
                     Jean-Marc Zucconi, Jeff Blomberg, Benny Halevy, Dave Boone,
                     Erik Habbinga, Kris Strecker, Walter Wong, Joshua Root,
                     Fabrice Bacchella, Zhenghua Xue, Qin Li, Darren Sawyer.
                     Ben England.

        Run began: Sun Dec 16 22:54:56 2012

        Include fsync in write timing
        Auto Mode
        File size set to 1024000 KB
        Record Size 4 KB
        Record Size 512 KB
        Record Size 16384 KB
        Excel chart generation enabled
        Command line used: iozone -e -a -s 1000M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2 -Rb chart.wks
        Output is in Kbytes/sec
        Time Resolution = 0.000001 seconds.
        Processor cache size set to 1024 Kbytes.
        Processor cache line size set to 32 bytes.
        File stride size set to 17 * record size.
                                                            random  random   
              KB  reclen   write rewrite    read    reread    read   write     
         1024000       4   14857   15165    20475    20480    3218     707
         1024000     512   14230   15890    18851    19415   17903    8245
         1024000   16384   13568   14394    19356    19965   20501   10173

iozone test complete.
So, now we have some "real world" numbers :)

Image
LICK FOR HIGH RES!


I've got some fancy stuff arriving for the Pi tomorrow, I'll re-run the tests and post the results :)

-Jamie M.
Seagate GoFlex Home, 1.2GHz ARM (kirkwood), 128MB RAM, Gigabit Ethernet, SATA2. Sandisk Extreme 120GB SSD running Arch ARM Linux 3.6.11-0. nginx + php-fpm = LIVE STATUS hosted right on the SGFH!! http://tafb.yi.org
User avatar
Posts: 136
Joined: Thu Dec 06, 2012 11:01 pm
by toysareforboys » Tue Dec 18, 2012 4:32 am
Ok, so not being very happy with my 4k random write performance, I went out and bought something a little faster :)

Image
LICK FOR HIGH RES!


The performance test with 1GB test file:

Code: Select all
pi@raspberrypi ~ $ sudo iozone -e -a -s 1000M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2
        Iozone: Performance Test of File I/O
                Version $Revision: 3.397 $
                Compiled for 32 bit mode.
                Build: linux

        Contributors:William Norcott, Don Capps, Isom Crawford, Kirby Collins
                     Al Slater, Scott Rhine, Mike Wisner, Ken Goss
                     Steve Landherr, Brad Smith, Mark Kelly, Dr. Alain CYR,
                     Randy Dunlap, Mark Montague, Dan Million, Gavin Brebner,
                     Jean-Marc Zucconi, Jeff Blomberg, Benny Halevy, Dave Boone,
                     Erik Habbinga, Kris Strecker, Walter Wong, Joshua Root,
                     Fabrice Bacchella, Zhenghua Xue, Qin Li, Darren Sawyer.
                     Ben England.

        Run began: Mon Dec 17 21:33:14 2012

        Include fsync in write timing
        Auto Mode
        File size set to 1024000 KB
        Record Size 4 KB
        Record Size 512 KB
        Record Size 16384 KB
        Command line used: iozone -e -a -s 1000M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2
        Output is in Kbytes/sec
        Time Resolution = 0.000001 seconds.
        Processor cache size set to 1024 Kbytes.
        Processor cache line size set to 32 bytes.
        File stride size set to 17 * record size.
                                                            random  random   
              KB  reclen   write rewrite    read    reread    read   write
         1024000       4   31018   27876    24220    24315    5069    8011
         1024000     512   30959   29528    23800    23996   23367   28443
         1024000   16384   30167   29145    24374    28547   30000   28942

iozone test complete.
Oh BABY! Now that's a little more like it :)

On to the charts!

The old SD card vs. the new SSD:

Image
LICK FOR HIGH RES!


Just the SSD graph:

Image
LICK FOR HIGH RES!


Works so so good :) WOOT!

-Jamie M.
Seagate GoFlex Home, 1.2GHz ARM (kirkwood), 128MB RAM, Gigabit Ethernet, SATA2. Sandisk Extreme 120GB SSD running Arch ARM Linux 3.6.11-0. nginx + php-fpm = LIVE STATUS hosted right on the SGFH!! http://tafb.yi.org
User avatar
Posts: 136
Joined: Thu Dec 06, 2012 11:01 pm
by neobusy » Tue Dec 18, 2012 10:47 pm
Currently running on my new 8GB card and I noticed horrible system stalls. Curious about it I came across this thread and gave the ever neglected Shizuku Edition a try :D She gave her best, but the results were... well... :?

Image

So I dug around and found a pretty old 1GB card and ran the tests again. Whoa!
After reading through some results here I ordered a Sandisk Ultra class 10 card - can't wait to try it!

Code: Select all
Kingston SDHC 8GB Class 4

           Sequential Read :    17.578 MB/s
          Sequential Write :     4.850 MB/s
         Random Read 512KB :    17.398 MB/s
        Random Write 512KB :     0.709 MB/s
    Random Read 4KB (QD=1) :     2.290 MB/s [   559.1 IOPS]
   Random Write 4KB (QD=1) :     0.006 MB/s [     1.5 IOPS]
   Random Read 4KB (QD=32) :     2.419 MB/s [   590.7 IOPS]
  Random Write 4KB (QD=32) :     0.012 MB/s [     3.0 IOPS]

  Test : 50 MB [R: 0.0% (0.0/7452.0 MB)] (x5)

Code: Select all
  Kingston 1GB 
 
           Sequential Read :     8.613 MB/s
          Sequential Write :     3.773 MB/s
         Random Read 512KB :     8.620 MB/s
        Random Write 512KB :     1.464 MB/s
    Random Read 4KB (QD=1) :     2.713 MB/s [   662.3 IOPS]
   Random Write 4KB (QD=1) :     0.023 MB/s [     5.6 IOPS]
   Random Read 4KB (QD=32) :     2.702 MB/s [   659.7 IOPS]
  Random Write 4KB (QD=32) :     0.026 MB/s [     6.4 IOPS]

  Test : 50 MB [R: 0.0% (0.0/970.1 MB)] (x5)
User avatar
Posts: 17
Joined: Mon Dec 03, 2012 11:24 am
by -heisenberg- » Wed Dec 19, 2012 8:37 am
I prefer bonnie for benchmarks, since it shows you more informative numbers like file create/delete, random seeks (very important) and CPU usage for each test. Just do

Code: Select all
sudo apt-get install bonnie++
cd ~
bonnie++


(Make sure you're on the actual file system of the device you want to test. Like your home dir for the internal SD card or the respective mount point in /media for a USB device.)

Here's what I got:

Code: Select all
SanDisk Cruzer Micro 8GB USB:

Version  1.96       ------Sequential Output------ --Sequential Input- --Random-
Concurrency   1     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
raspberrypi      1G    84  96  5609   8  4472  11   567  99 26954  25  33.2   4
Latency              1349ms    2010ms    6570ms   33646us   17988us   21615us
Version  1.96       ------Sequential Create------ --------Random Create--------
raspberrypi         -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                 16  1047  11 +++++ +++  3388  25  3145  38 +++++ +++  3767  28
Latency              1852us    4302us    4252us    1812us     174us    1646us


Samsung MB-MS2G (2GB Micro SD card) in SD card slot:

Version  1.96       ------Sequential Output------ --Sequential Input- --Random-
Concurrency   1     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
raspberrypi      1G    84  96  7542   8  5223   9   559  99 19380  14  77.0   7
Latency               563ms   10308ms    4605ms   34502us    9243us   25281us
Version  1.96       ------Sequential Create------ --------Random Create--------
raspberrypi         -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                 16  3976  44 +++++ +++  6630  50  4370  53 +++++ +++  4750  43
Latency              1824us    5327us    4264us    2503us     225us     150ms


SanDisk Mobile Ultra SDHC 4GB in SD card slot:

Version  1.96       ------Sequential Output------ --Sequential Input- --Random-
Concurrency   1     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
raspberrypi      1G    87  95 18742  19  9842  17   647  99 25127  17 144.8  13
Latency               520ms    3623ms     700ms   27117us   22523us   42829us
Version  1.96       ------Sequential Create------ --------Random Create--------
raspberrypi         -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                 16  4772  57 +++++ +++  7133  57  5207  66 +++++ +++  6817  59
Latency              6853us    4337us    4153us     817us      91us     536us
arm_freq=1100 / core_freq=500 / sdram_freq=450 / over_voltage=6
Posts: 21
Joined: Mon Dec 17, 2012 7:12 am
by okrasz » Thu Dec 27, 2012 2:48 pm
PRETEC 16 GB Class 10
--------------------------------------------------------------------
CrystalDiskMark 3.0.2 (C) 2007-2012 hiyohiyo
Crystal Dew World : http://crystalmark.info/
--------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 20.913 MB/s
Sequential Write : 12.257 MB/s
Random Read 512KB : 20.597 MB/s
Random Write 512KB : 1.125 MB/s
Random Read 4KB (QD=1) : 4.230 MB/s [ 1032.8 IOPS]
Random Write 4KB (QD=1) : 0.018 MB/s [ 4.4 IOPS]
Random Read 4KB (QD=32) : 4.763 MB/s [ 1162.8 IOPS]
Random Write 4KB (QD=32) : 0.016 MB/s [ 3.9 IOPS]

Test : 50 MB [G: 58.9% (8.9/15.0 GB)] (x2)
Date : 2012/12/27 15:28:23
OS : Windows 7 Professional SP1 [6.1 Build 7601] (x64)


Kingston 2GB Class unknown
-----------------------------------------------------------------------
CrystalDiskMark 3.0.2 (C) 2007-2012 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 15.721 MB/s
Sequential Write : 2.829 MB/s
Random Read 512KB : 15.550 MB/s
Random Write 512KB : 0.598 MB/s
Random Read 4KB (QD=1) : 3.243 MB/s [ 791.7 IOPS]
Random Write 4KB (QD=1) : 0.006 MB/s [ 1.5 IOPS]
Random Read 4KB (QD=32) : 3.496 MB/s [ 853.6 IOPS]
Random Write 4KB (QD=32) : 0.006 MB/s [ 1.6 IOPS]

Test : 50 MB [G: 58.3% (1082.3/1857.7 MB)] (x2)
Date : 2012/12/27 15:41:22
OS : Windows 7 Professional SP1 [6.1 Build 7601] (x64)

SandDisk Ultra II Class 4 15 MB/s
-----------------------------------------------------------------------
CrystalDiskMark 3.0.2 (C) 2007-2012 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 18.226 MB/s
Sequential Write : 18.000 MB/s
Random Read 512KB : 18.022 MB/s
Random Write 512KB : 3.154 MB/s
Random Read 4KB (QD=1) : 5.384 MB/s [ 1314.5 IOPS]
Random Write 4KB (QD=1) : 0.035 MB/s [ 8.6 IOPS]
Random Read 4KB (QD=32) : 5.725 MB/s [ 1397.6 IOPS]
Random Write 4KB (QD=32) : 0.035 MB/s [ 8.5 IOPS]

Test : 50 MB [G: 80.4% (3118.1/3878.5 MB)] (x2)
Date : 2012/12/27 15:47:55
OS : Windows 7 Professional SP1 [6.1 Build 7601] (x64)
Posts: 1
Joined: Thu Dec 27, 2012 2:27 pm
by toysareforboys » Thu Dec 27, 2012 3:52 pm
Yikes. It's that 4k random write that's the killer :(

I just couldn't take the horrible i/o performance on my Pi any more, so I replaced it with something a little stranger :) http://tafb.yi.org

iozone i/o performance, look at the 4k read/writes! :D

Image
LICK FOR HIGH RES!


Oh baby, SO FAST :D

-Jamie M.
Last edited by toysareforboys on Thu Dec 27, 2012 4:18 pm, edited 1 time in total.
Seagate GoFlex Home, 1.2GHz ARM (kirkwood), 128MB RAM, Gigabit Ethernet, SATA2. Sandisk Extreme 120GB SSD running Arch ARM Linux 3.6.11-0. nginx + php-fpm = LIVE STATUS hosted right on the SGFH!! http://tafb.yi.org
User avatar
Posts: 136
Joined: Thu Dec 06, 2012 11:01 pm
by milhouse » Thu Dec 27, 2012 4:03 pm
I think it's been asked before, but what is the point of comparing SD/SSD performance in a massively more powerful Windows PC when the storage device will ultimately be used with a Pi? Surely your time would be better spent running benchmarks using your Pi?
Posts: 552
Joined: Mon Jan 16, 2012 12:59 pm
by toysareforboys » Thu Dec 27, 2012 4:15 pm
milhouse wrote:I think it's been asked before, but what is the point of comparing SD/SSD performance in a massively more powerful Windows PC when the storage device will ultimately be used with a Pi? Surely your time would be better spent running benchmarks using your Pi?
Who's that comment directed at? Me? Who's running benchmarks under windows? Did you click the link or look at the picture even? Windows? LOL!

My SSD test on the Pi was linked a little earlier in this thread :)

-Jamie M.
Seagate GoFlex Home, 1.2GHz ARM (kirkwood), 128MB RAM, Gigabit Ethernet, SATA2. Sandisk Extreme 120GB SSD running Arch ARM Linux 3.6.11-0. nginx + php-fpm = LIVE STATUS hosted right on the SGFH!! http://tafb.yi.org
User avatar
Posts: 136
Joined: Thu Dec 06, 2012 11:01 pm
by milhouse » Thu Dec 27, 2012 4:22 pm
toysareforboys wrote:Who's that comment directed at? Me? Who's running benchmarks under windows? Did you click the link or look at the picture even? Windows? LOL!


It was a general question, so not you specifically, or anyone specifically for that matter, apologies if it read as though it was directed at an individual. I'd actually looked at the post above yours which is using Windows to test, and I'm not really sure what relevance this will have when using a Pi.
Posts: 552
Joined: Mon Jan 16, 2012 12:59 pm
by toysareforboys » Thu Dec 27, 2012 4:31 pm
milhouse wrote:It was a general question, so not you specifically, or anyone specifically for that matter, apologies if it read as though it was directed at an individual. I'd actually looked at the post above yours which is using Windows to test, and I'm not really sure what relevance this will have when using a Pi.
Ahhh, ok :) When no quote button is clicked one must assume :D

-Jamie M.
Seagate GoFlex Home, 1.2GHz ARM (kirkwood), 128MB RAM, Gigabit Ethernet, SATA2. Sandisk Extreme 120GB SSD running Arch ARM Linux 3.6.11-0. nginx + php-fpm = LIVE STATUS hosted right on the SGFH!! http://tafb.yi.org
User avatar
Posts: 136
Joined: Thu Dec 06, 2012 11:01 pm
by iwl » Mon Jan 21, 2013 5:30 pm
Have just tested an (good on small random I/O, Pi also feels fast):
Samsung MB-SSBGA Class 10 Essential SDHC 32GB
Code: Select all
./iozone -e -I -a -s 50M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2
                                                     random  random
       KB  reclen   write rewrite    read    reread    read   write
    51200       4    1809    1893     6283     6272    6225    1078
    51200     512   10721   11266    21952    21903   21897   11206
    51200   16384   11332   11333    22438    22427   22435   11386

Sandisk Extreme 45MB/s 32 GB slower on small random (I think is important) only larger writes are faster and is more expensive
Code: Select all
                                                      random  random
       KB  reclen   write rewrite    read    reread    read   write
    51200       4    1844    1947     4872     4521    3544     739
    51200     512   20620   20880    21852    21850   21644    4246
    51200   16384   16088   20402    22336    22338   22311   20317
Posts: 16
Joined: Mon Dec 17, 2012 5:35 pm
by toysareforboys » Mon Jan 21, 2013 6:18 pm
iwl wrote:Have just tested an (good on small random I/O, Pi also feels fast):
Samsung MB-SSBGA Class 10 Essential SDHC 32GB
Code: Select all
./iozone -e -I -a -s 50M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2
                                                     random  random
       KB  reclen   write rewrite    read    reread    read   write
    51200       4    1809    1893     6283     6272    6225    1078
    51200     512   10721   11266    21952    21903   21897   11206
    51200   16384   11332   11333    22438    22427   22435   11386
As per the I/O zone manual "you should pick a test file size TWICE the size of your available RAM, to ensure a proper benchmark".

I would be interested to see your Samsung card tested on one of the following lines (depending on your ram/model Pi):

For 512mb Pi try: ./iozone -e -I -a -s 1000M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2
Or 256mb Pi try: ./iozone -e -I -a -s 500M -r 4k -r 512k -r 16M -i 0 -i 1 -i 2

-Jamie M.
Seagate GoFlex Home, 1.2GHz ARM (kirkwood), 128MB RAM, Gigabit Ethernet, SATA2. Sandisk Extreme 120GB SSD running Arch ARM Linux 3.6.11-0. nginx + php-fpm = LIVE STATUS hosted right on the SGFH!! http://tafb.yi.org
User avatar
Posts: 136
Joined: Thu Dec 06, 2012 11:01 pm
by iwl » Mon Jan 21, 2013 7:13 pm
You can clearly See that the Test is plausible small sectors are Slower, random is Slower, writing is Slower. The switches are Set for disabling Caching / Buffering.
And I can Feel the pi is more responsive with These card. I dont like to wear my card that much.
Posts: 16
Joined: Mon Dec 17, 2012 5:35 pm
by toysareforboys » Mon Jan 21, 2013 7:33 pm
iwl wrote:You can clearly See that the Test is plausible small sectors are Slower, random is Slower, writing is Slower. The switches are Set for disabling Caching / Buffering.
And I can Feel the pi is more responsive with These card. I dont like to wear my card that much.
Yeah, but when I see your random write (11206) is faster than your sequential write (10721), something must be wrong ;) There must be some caching going on somewhere :( Without the large test file, it's impossible to tell.

-Jamie M.
Seagate GoFlex Home, 1.2GHz ARM (kirkwood), 128MB RAM, Gigabit Ethernet, SATA2. Sandisk Extreme 120GB SSD running Arch ARM Linux 3.6.11-0. nginx + php-fpm = LIVE STATUS hosted right on the SGFH!! http://tafb.yi.org
User avatar
Posts: 136
Joined: Thu Dec 06, 2012 11:01 pm