SD Strange-results - or - How I learned to love CM7 on SD - Dec-11 results compiled!

Search This thread

kkt12

Member
Sep 14, 2012
46
11
The Samsung class 10 has a competitor, the SanDisk Mobile Ultra UHS-I, class 10 microSD!

SanDisk's description of the card

Benchmarks for the 32GB, item# SDSDQUA-032G-A11A, USB 2.0 reader connected to USB 3.0 port
Code:
           Sequential Read :    18.372 MB/s
          Sequential Write :    13.057 MB/s
         Random Read 512KB :    18.703 MB/s
        Random Write 512KB :     8.554 MB/s
    Random Read 4KB (QD=1) :     3.150 MB/s [   769.0 IOPS]
   Random Write 4KB (QD=1) :     1.957 MB/s [   477.8 IOPS]
   Random Read 4KB (QD=32) :     3.590 MB/s [   876.4 IOPS]
  Random Write 4KB (QD=32) :     1.960 MB/s [   478.5 IOPS]

  Test : 50 MB [D: 0.0% (0.0/29.7 GB)] (x5)


this is indeed a great card. I am getting similarly high scores. it's on sale right now on Newegg (I see your review there!), 15% off with included promo code.


http://www.newegg.com/Product/Produ...Micro SDHC Flash Card Model SDSDQUA-032G-A11A
 

bastard_hun

New member
Feb 22, 2008
2
0
QUMOX 32GB Class 6 Micro SD SDHC MicroSD Memory Card 32 GB

I ordered it from ebay before I even saw this topic, so it was a gamble. It was approx 25$

Fortunately luck was on my side :)

These are the results with a really crappy chinese card reader:

-----------------------------------------------------------------------
CrystalDiskMark 3.0.2 x64 (C) 2007-2013 hiyohiyo
Crystal Dew World :
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 16.320 MB/s
Sequential Write : 7.298 MB/s
Random Read 512KB : 16.469 MB/s
Random Write 512KB : 12.371 MB/s
Random Read 4KB (QD=1) : 4.735 MB/s [ 1156.0 IOPS]
Random Write 4KB (QD=1) : 1.173 MB/s [ 286.3 IOPS]
Random Read 4KB (QD=32) : 5.969 MB/s [ 1457.3 IOPS]
Random Write 4KB (QD=32) : 1.599 MB/s [ 390.3 IOPS]

Test : 50 MB [E: 0.0% (0.0/29.7 GB)] (x5)
Date : 2013/02/09 9:16:08
OS : Windows Server 2008 R2 Enterprise Edition (Full installation) SP1 [6.1 Build 7601] (x64)
 

Socket

Member
Mar 22, 2003
26
1
High failure rate on Sandisk Mobile Ultra 20SDSDQUA-032G-A11A?

Based on this post I just ordered this card (from Amazon). Hasn't arrived yet.

I just now looked through the reviews over on Newegg, using the link below. Soobaerodude's review there was one of the first, posted back in May 2012 and things looked rosy then. But since that time, there have been a LOT (like about half of all the reviews) that indicate this card either failed immediately, or failed within 2-3 months

I'd be interested to hear from people here who bought this specific card (20SDSDQUA-032G-A11A) and have more than 2 months under their belt. Is it working ok? Soobaerodude, how about you? I'm worried I'm heading for the same problems...

Thanks


this is indeed a great card. I am getting similarly high scores. it's on sale right now on Newegg (I see your review there!), 15% off with included promo code.


http://www.newegg.com/Product/Produ...Micro SDHC Flash Card Model SDSDQUA-032G-A11A
 

Socket

Member
Mar 22, 2003
26
1
Sandisk 32GB Ultra UHS-1 MicroSD Card arrived, here are my initial performance testing results.
5 iterations, 50 MB test file.

-----------------------------------------------------------------------
CrystalDiskMark 3.0.2 x64 (C) 2007-2013 hiyohiyo
Crystal Dew World : http://crystalmark.info/

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 21.722 MB/s
Sequential Write : 12.715 MB/s
Random Read 512KB : 21.309 MB/s
Random Write 512KB : 12.037 MB/s
Random Read 4KB (QD=1) : 5.631 MB/s [ 1374.8 IOPS]
Random Write 4KB (QD=1) : 1.837 MB/s [ 448.5 IOPS]
Random Read 4KB (QD=32) : 5.163 MB/s [ 1260.5 IOPS]
Random Write 4KB (QD=32) : 2.091 MB/s [ 510.5 IOPS]

Test : 50 MB [M: 0.0% (0.0/29.7 GB)] (x5)
Date : 2013/02/24 13:33:39
OS : Windows 7 Home Premium Edition [6.1 Build 7600] (x64)
-----------------------------------------------------------------------

This was run using a USB3.0 port on my PC but with a USB2.0 to MicroSD adapter (the compact keychain type).

I also ran 3 other benchmarks using other connection methods like the included Sandisk SD-to-MicroSD card adapter inserted into my PC's built-in memory card reader, using a Kingston SD-to-MicroSD card adapter in that same PC reader, etc. No single test run was consistently better than the others considering all 8 test results.. some where higher, some were lower. The one I was most interested in -- Random Write 4KB (QD=32) -- varied from 1.783 to 2.195.

We'll see how much better CMv10-on-ext-SD works for me on that card once I get my current 8G card content migrated over to it.

Also, I guess I'll just have to wait and see if this card will die on me in a month or two like many reviews on newegg stated, or will last longer.


Based on this post I just ordered this card (from Amazon). Hasn't arrived yet.

Thanks
 

Socket

Member
Mar 22, 2003
26
1
Sandisk 32GB Ultra UHS-1 MicroSD Card arrived, here are my initial performance testing results.
[...]
We'll see how much better CMv10-on-ext-SD works for me on that card once I get my current 8G card content migrated over to it.

Update: I migrated my 8GB card content over to the Sandisk 32G Ultra UHS-1, and it definitely runs much faster than the 8 GB card, as one would expect given the much higher 4k random write results.

However this card, when running CM10 from it, either won't boot properly, or if it does, the session doesn't last long before spontaneously rebooting (probably due to kernel panic). I even rebuilt it from the ground up (the first migration was by cloning the disk image). Both attempts resulted in unstable booting and running. When I switch back to the old 8GB card with the same content on it, CM10 is stable and fine.

So it unfortunately looks like this model of card does have a reliability issue, and I didn't have to wait 2 months either.. it was having issues immediately on the first day. If I can find a Windows application that lets me "realign" ext partitions, I'll make another attempt at this. But if not, this card is getting returned.
 

leapinlar

Senior Member
Oct 18, 2006
8,873
3,878
Update: I migrated my 8GB card content over to the Sandisk 32G Ultra UHS-1, and it definitely runs much faster than the 8 GB card, as one would expect given the much higher 4k random write results.

However this card, when running CM10 from it, either won't boot properly, or if it does, the session doesn't last long before spontaneously rebooting (probably due to kernel panic). I even rebuilt it from the ground up (the first migration was by cloning the disk image). Both attempts resulted in unstable booting and running. When I switch back to the old 8GB card with the same content on it, CM10 is stable and fine.

So it unfortunately looks like this model of card does have a reliability issue, and I didn't have to wait 2 months either.. it was having issues immediately on the first day. If I can find a Windows application that lets me "realign" ext partitions, I'll make another attempt at this. But if not, this card is getting returned.
Just to remind folks, he's running on an HD+ so what he said about instability may not apply to Nook Colors.

Sent from my Nook HD+ running CM10 on Hybrid SD
 

Socket

Member
Mar 22, 2003
26
1
Just to remind folks, he's running on an HD+ so what he said about instability may not apply to Nook Colors.

Right, this was testing on my Nook HD+.

However as it turns out, I also have a Nook Color which I set up a few months ago for dual boot, with CM7 on external SD (another 8GB slow card). It's stable using that 8G card.

I think I'll clone that card content over to this 32G Sandisk and see if the 32G card causes instability on a Nook Color/CM7 also. Stay tuned...
 
Last edited:

Socket

Member
Mar 22, 2003
26
1
Right, this was testing on my Nook HD+.

However as it turns out, I also have a Nook Color which I set up a few months ago for dual boot, with CM7 on external SD (another 8GB slow card). It's stable using that 8G card.

I think I'll clone that card content over to this 32G Sandisk and see if the 32G card causes instability on a Nook Color/CM7 also. Stay tuned...

Update:

I cloned my Nook color 8g Cm7 ext SD card over to this 32G Sandisk (I didn't extend any partitions, so it's just got the same 8GB of partitions on it) and tried it in the Nook color.

No dice - totally unstable. Couldn't even boot. I would get the "Cyanogen.. Loading.." boot splash screen for 30 seconds, then it would go dark, then repeat, then go dark, then stayed dark permanently. I manually rebooted.. .same thing 2nd time. Popped the old 8GB CM7 card back in, and it booted fine.

So at this point I think I can definitively say that this Sandisk 32GB Ultra UHC-1 Class 10 card is total crap, both for Nook Color w CM7 on SD, and Nook HD+ with CM10 on SD.
 

tonestertm

Member
Sep 21, 2011
20
10
Sorry, I don't think you can "definitively" say anything about that card with Nook Colors yet. If you start from scratch with proper image, partitioning and an MD5 checked ROM, (no cloning, and only migrating data) and it still won't go, THEN you can write it off. :)

Sent from my MOTWX435KT using Tapatalk 2
 
Last edited:

Socket

Member
Mar 22, 2003
26
1
Sorry, I don't think you can "definitively" say anything about that card with Nook Colors yet. If you start from scratch with proper image, partitioning and an MD5 checked ROM, (no cloning, and only migrating data) and it still won't go, THEN you can write it off. :)

Sent from my MOTWX435KT using Tapatalk 2

I've put all the effort into the nook color side that I planned to. I'm satisfied with my conclusion as far as my needs go.

If someone else wants to run an experiment on nook color rebuilding this particular model card from scratch instead of cloning, they are welcome to do so.
 

leapinlar

Senior Member
Oct 18, 2006
8,873
3,878
I've put all the effort into the nook color side that I planned to. I'm satisfied with my conclusion as far as my needs go.

If someone else wants to run an experiment on nook color rebuilding this particular model card from scratch instead of cloning, they are welcome to do so.
You must have a bad card. I use these on my Nook Colors all the time with no issues.

Sent from my Nook HD+ running CM10 on Hybrid SD
 

wrevilo

Member
Mar 7, 2013
5
2
I hope to get my sister's unused Nook Color to play around with and will stick CM10 upon it. I've benchmarked a couple of micro SD cards I have in other devices - an 8Gb Samsung in my Nook Simple Touch and a 32Gb Sandisk in my Sandisk Sansa Clip+. Here are the results:
Samsung 8Gb Class 10 MB-MP8GA
CrystalDiskMark 3.0.2 x64 (C) 2007-2013 hiyohiyo
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]
Sequential Read : 19.219 MB/s
Sequential Write : 17.694 MB/s
Random Read 512KB : 18.535 MB/s
Random Write 512KB : 2.302 MB/s
Random Read 4KB (QD=1) : 2.511 MB/s [ 613.1 IOPS]
Random Write 4KB (QD=1) : 0.022 MB/s [ 5.3 IOPS]
Random Read 4KB (QD=32) : 2.698 MB/s [ 658.6 IOPS]
Random Write 4KB (QD=32) : 0.028 MB/s [ 6.9 IOPS]
Test : 50 MB [E: 0.5% (36.5/7687.0 MB)] (x5)
Date : 2013/03/07 8:48:31
OS : Windows 7 Home Premium Edition SP1 [6.1 Build 7601] (x64)
Sandisk 32Gb Class 2
CrystalDiskMark 3.0.2 x64 (C) 2007-2013 hiyohiyo
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]
Sequential Read : 12.655 MB/s
Sequential Write : 5.757 MB/s
Random Read 512KB : 12.860 MB/s
Random Write 512KB : 5.348 MB/s
Random Read 4KB (QD=1) : 0.913 MB/s [ 222.9 IOPS]
Random Write 4KB (QD=1) : 2.015 MB/s [ 492.0 IOPS]
Random Read 4KB (QD=32) : 0.630 MB/s [ 153.7 IOPS]
Random Write 4KB (QD=32) : 2.136 MB/s [ 521.4 IOPS]
Test : 50 MB [E: 53.3% (15.8/29.7 GB)] (x5)
Date : 2013/03/07 9:12:43
OS : Windows 7 Home Premium Edition SP1 [6.1 Build 7601] (x64)
Both tested via a Samsung adapter in my laptop's SD card reader.

As you can see the Sandisk dominates the random write tests. I think I will pick up a 32Gb Sandisk or Samsung, test them and if the random write is significantly worse than my existing Sandisk I will swap some data around to get the best card for the job in the Nook Color.
 

Socket

Member
Mar 22, 2003
26
1
[...] As you can see the Sandisk dominates the random write tests. I think I will pick up a 32Gb Sandisk or Samsung, test them and if the random write is significantly worse than my existing Sandisk I will swap some data around to get the best card for the job in the Nook Color.

Be careful, my 32GB Sandisk Ultra (class 10) is unstable. I recall reading in this forum that others had stability issues with Sandisk cards over 16 GB.

I just received two Sandisk 16 GB Class 4 cards and am in the process of burning CM 10 to one of them. Here are the CrystalDiskMark stats for one of them:

-----------------------------------------------------------------------
CrystalDiskMark 3.0.2 x64 (C) 2007-2013 hiyohiyo
Crystal Dew World :
http://crystalmark.info/

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 19.373 MB/s
Sequential Write : 5.863 MB/s
Random Read 512KB : 19.030 MB/s
Random Write 512KB : 3.376 MB/s
Random Read 4KB (QD=1) : 4.323 MB/s [ 1055.4 IOPS]
Random Write 4KB (QD=1) : 1.471 MB/s [ 359.2 IOPS]
Random Read 4KB (QD=32) : 4.542 MB/s [ 1109.0 IOPS]
Random Write 4KB (QD=32) : 1.333 MB/s [ 325.3 IOPS]

Test : 50 MB [M: 0.0% (0.0/14.8 GB)] (x5)
Date : 2013/03/10 12:28:34
OS : Windows 7 Home Premium Edition [6.1 Build 7600] (x64)
 
Last edited:

wrevilo

Member
Mar 7, 2013
5
2
OK so I received my 32GB Sandisk Ultra class 10 card and have tested it:
Sandisk 32Gb Ultra
CrystalDiskMark 3.0.2 x64 (C) 2007-2013 hiyohiyo
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]
Sequential Read : 19.306 MB/s
Sequential Write : 11.043 MB/s
Random Read 512KB : 18.804 MB/s
Random Write 512KB : 12.216 MB/s
Random Read 4KB (QD=1) : 2.909 MB/s [ 710.2 IOPS]
Random Write 4KB (QD=1) : 1.486 MB/s [ 362.7 IOPS]
Random Read 4KB (QD=32) : 3.182 MB/s [ 776.9 IOPS]
Random Write 4KB (QD=32) : 1.691 MB/s [ 412.9 IOPS]
Test : 50 MB [E: 0.0% (0.0/29.7 GB)] (x5)
Date : 2013/03/13 12:33:51
OS : Windows 7 Home Premium Edition SP1 [6.1 Build 7601] (x64)
It is now in the NC and will be used as a media drive, as I installed CM10 to the internal memory. Shame it's not as fast as the Samsung for sequential writes, which would have been better for writing files over from my desktop.
 

DonEdwards

New member
May 19, 2013
4
1
An answer, and a surprising slight problem...

Answer first. A while back there was some wondering why it's small-block random *write* speed that seems to be critical, but random *read* speed doesn't seem to matter so much. The theorizing was that it's plenty fast enough in all cases, and maybe if we found a device where random read speed is slower than random write speed, we could draw some better conclusions.

Well, I found such a device.

4K random, Q=1: read 0.413, write 1.572
4K random, Q=32: read 0.903, write 1.806

Those random read speeds are about the worst we've seen in this thread. While the random write speeds are near the top.

I don't know how well this device would work for Android, but it's just fine as the boot (and primary storage, and swapfile) device for Windows 7 64-bit. And it works pretty good for Linux too, according to quite a few people.

Yep, it's the HD (complete with spinning disk) in my laptop.

A Toshiba USB drive is essentially the same speed at Q=1, but about 2/3 that speed at Q=32. So these numbers are probably typical of HDs of similar age.

So, it looks like "everything out there is plenty fast enough in that category" is almost certainly the correct answer.

Full results with 5 repetitions, just for the sake of completeness:

Samsung HM250HI 250GB hard drive, internal
Sequential read 69.43 write 67.54
Random512K read 30.09 write 41.70
Random4K,Q=1 read 0.413 write 1.572
Random4K,Q=32 read 0.903 write 1.806

Toshiba 500GB USB3 drive
Sequential read 29.36 write 18.53
Random512K read 19.92 write 23.50
Random4K,Q=1 read 0.427 write 1.362
Random4K,Q=32 read 0.638 write 1.355

----

Now for the surprising slight problem: I've observed that CrystalDiskMark 3.0.2 (can't comment on any earlier version) does NOT average the results of multiple tests. It shows the best single-test result. Next time you're running it, actually sit and watch what it puts in the result boxes... in each box it puts the result of EACH test run as it finishes, until it finishes the last of however-many-you're-doing when it puts up either the highest number it has shown for one of this batch or a new number that is even higher.

If it were averaging the results, then the more times you ran the test the more *typical* and *reliable* your result would be. But with it instead showing the best result, the more times you run the test the more *exceptional* and *unreliable* your result is likely to be.

In the times I was watching it, I frequently saw that the lowest result in a 5-repetition cycle would be around half of the final result. (And of course I never saw a result higher than the final result, or I wouldn't be talking about this.) So there's obviously plenty of room for exceptional results.
 

Agwalters

New member
May 19, 2013
1
0
Thank you!

1/16/2012 update: Thanks to waxhell for compiling all of the results in this thread (at least as of mid-december! OK, so I took a while to get this posted...)
https://docs.google.com/spreadsheet/ccc?key=0AjPE3ZAD2eVudE9vZmQ3aHlfTkFvU2J4ZUplRDJQTEE

5/20 update: SD card performance benchmark table by a.fenderson from later in this thread added at the bottom of this post. Thank you a.fenderson!
------------------
April minor update with a Transcend SD (see http://forum.xda-developers.com/showpost.php?p=12964262&postcount=8)

_________________________________________________

Let me tell you my story.

Over the last couple of months as I have mucked around with SD booting various flavors of froyo and, more recently, CM7, I have found it maddening that there are apparently so many people that LOVE those versions, even people that seemed to have the same SD card as the two I've tried.

One of those is a Class 10 Patriot 8gig, which I've seen mentioned as an acceptable boot disk.
The other is a Kingston Class 4 4gig that tests out as significantly slower than the Patriot, but didn't really run froyo any slower.

Now I know I've seen posts that mention small block write speed as being important, but the numbers I've seen posted really didn't make me think the Patriot was the cause...
...the cause of FRUSTRATINGLY slow UI, where froyo (and CM7) seem to go off into lala-land for a few seconds every time I (tried to) do something.
...the cause of glacial web-surfing, where even downloading GOOGLEforchrissake takes forever. Of snail's-pace market downloads. Oh god.

Every time I would give up and go back to Eclair and breath a sigh of relief, reveling in the snappiness, the zippy web-surfing, the rapid response of m.pornhub.com.

And every time the siren song of the CM7-elite would call me back. THIS time I did something different. I saw mention that a Class 2, YES A FREAKING CLASS 2, Sandisk from costco worked well. So I skippity-skopped up and bought one.

After backing up the Patriot and writing THAT VERY SAME img to the Sandisk (which, yes, took 3 times as long as writing to the Patriot), I booted CM7.

OMG

The heavens opened and the angels sang, I HAD ARRIVED IN THE PROMISED LAND!
NOW I knew what the chosen had been praising! NOW I knew the joy that is CM7! ZIP-ZIP-ZIPPITY!

HOLY CRAP?! How could a benchmark be so misleading? I HATE those bastard SD manufacturers, or maybe it's the industry group that chose such a sucky measure of speed.

For your edification, here are some CrystalDiskMark (5 iterations, 50MB) results from 2 sucky cards and the good Sandisk.


__Test_________________Sandisk 8G Class 2___Patriot 8G Class 10____Kingston Class4 4Gig
Sequential Read :__________10.871 MB/s________20.036 MB/s___________18.700 MB/s
Sequential Write :__________6.659 MB/s________13.660 MB/s____________4.277 MB/s
Random Read 4KB (QD=32) :___3.077 MB/s_________3.444 MB/s____________2.088 MB/s
Random Write 4KB (QD=32) :__1.791 MB/s________0.038 MB/s_____________0.016 MB/s


Look at that, both the Kingston and the Patriot blow away or are close enough to the Sandisk in everything BUT....

(wait for it)....

SMALL BLOCK RANDOM WRITES
where the Sandisk is FIFTY times the speed of the Patriot and ONE HUNDREDtimes the Kingston.

Now you may say, "Swizzlenuts, old pal, I KNEW that."

But for all you poor slobs who didn't, who are dragging yer sorry ass through the broken glass that is booting off of a slow SD (you know who both of you are), I hope this helps.

And now you may commence posting links to specific posts where this info was discussed in detail last December.


------------------
Thanks to all of the people that posted their results and to a.fenderson for compiling them. Here's his table from http://forum.xda-developers.com/showpost.php?p=13991898&postcount=144
Code:
POSTER         BRAND      CLASS  CAPACITY  4 KB RANDOM
                                  (GB)    WRITE QD32(MB/s)    NOTE
swoozle        SanDisk        2    8          1.791           Model: SDSQ-8192-AC11M
a.fenderson    SanDisk        4    16         1.660           SanDisk C4 16GB 1 of 2 cards
MickMcGeough   SanDisk        4    8          1.59            benchmarked in XBench, QD unknown
a.fenderson    SanDisk        4    16         1.500           SanDisk C4 16GB 2 of 2 cards
Awats          SanDisk        4    16         1.391           SanDisk C4 16GB 1 of 2 cards ; fastest (onboard) reader
arwild01       SanDisk        4    8          1.369    
Awats          SanDisk        4    16         1.284           SanDisk C4 16GB 2 of 2 cards ; fastest (onboard) reader
robot8         SanDisk        4    8          1.270           SanDisk C4 8GB 2 of 2 cards
robot8         SanDisk        4    8          1.259           SanDisk C4 8GB 1 of 2 cards
chinly43       SanDisk        4    16         1.257           (not via Nook) ; same card as chinly43's other listed SanDisk C4 16GB
joobu          SanDisk        4    4          1.175    
angomy         Nook internal  N/A N/A         1.116           via Nook on USB
chinly43       Nook internal  N/A N/A         1.094           via Nook on USB
a.fenderson    Transcend      2    32         1.032    
a.fenderson    SanDisk        4    4          0.898    
a.fenderson    SanDisk        2    4          0.891    
pchoi94        SanDisk        4    16         0.834    
chinly43       SanDisk        4    16         0.769           via Nook on USB ; same as chinly43's other listed SanDisk C4 16GB
pryonix        SanDisk        4    8          0.625    
chinly43       SanDisk        2    8          0.616    
a.fenderson    SanDisk        4    8          0.596    
angomy         SanDisk        4    16         0.574    
Blue6IX        SanDisk        2    16         0.350    
a.fenderson    SanDisk       N/A   2          0.269    
Awats          SanDisk        4    2          0.261           SanDisk C4 2GB 1 of 2 cards ; fastest (onboard) reader
Awats          SanDisk        4    2          0.236           SanDisk C4 2GB 2 of 2 cards ; fastest (onboard) reader
pryonix        Samsung       N/a   2          0.093    
victle         Kingston      N/A   2          0.051    
victle         Dane-Elec     N/A   2          0.050           0.050 or less:  exact value unspecified
joobu          Lexar          4    8          0.038    
swoozle        Patriot        10   8          0.038    
Blue6IX        Dane-Elec      4    4          0.037           (made in Japan)
omghahalol     Transcend      6    16         0.037    
Ravynmagi      Samsung(??)    2    4          0.037    
pryonix        Kingmax        10   16         0.036    
Ravynmagi      Wintech        10   16         0.036    
victle         Transcend      6    4          0.034    
omghahalol     Transcend      6    8          0.033    
Ravynmagi      SanDisk        2    8          0.033    
pryonix        Transcend      6    8          0.033    
Ravynmagi      Patriot        10   16         0.030    
Blue6IX        PNY            10   8          0.030    
pchoi94        Kingston       2    16         0.030    
swoozle        Transcend      6    8          0.029           Model: TS8GUSDHC6
joobu          ??            N/A   2          0.029    
chinly43       SanDisk       N/A   1          0.029    
a.fenderson    SanDisk       N/A   1          0.028    
victle         PNY            4    8          0.028    
Blue6IX        PNY           N/A   2          0.027           (made in Taiwan)
omghahalol     SanDisk        2    2          0.021    
MickMcGeough   SanDisk        2    8          0.02           benchmarked in XBench, QD unknown
robot8         Transcend      6    8          0.018           Transcend C6 8GB 1 of 2 cards
ExploreMN      Patriot        10   16         0.018    
arwild01       Samsung(??)    4    8          0.017    
swoozle        Kingston       4    4          0.016           Model: SDC4/8GB
Awats          Patriot        4    4          0.016           Patriot C4 4GB 1 of 2 cards ; fastest (onboard) reader
robot8         Transcend      6    8          0.014           Transcend C6 8GB 2 of 2 cards
Tnexus         Patriot        10   16         0.014    
a.fenderson    Kingston       4    8          0.014    
Awats          Patriot        4    4          0.011           Patriot C4 4GB 2 of 2 cards ; fastest (onboard) reader
chinly43       Lexar          4    8          0.011

Thank you for this post. I had rooted my Nook color using a Sandisk 32 GB class 10 SD card only to find the delay time intolerable. Used 8GB class 2 and the Heavens parted! Now running CM 10.1 with no problems!
 

tonestertm

Member
Sep 21, 2011
20
10
I've observed that CrystalDiskMark 3.0.2 (can't comment on any earlier version) does NOT average the results of multiple tests. It shows the best single-test result...
...
In the times I was watching it, I frequently saw that the lowest result in a 5-repetition cycle would be around half of the final result....
Don, I'm glad you posted that, because that's been bugging me for a long time, but I decided to just ignore it and consider the results as "relative". Since everything I'm doing with my cards is on the same test bed, and I tend to run the test multiple times, watching the results along the way (BTW, I think the lowest "intermediate" score I've seen was maybe 2/3 the top speed, I've never had anything as low as 1/2), I know approximately how each card compares to my others. As far as comparison to others in this thread, well, there are so many variables anyway - readers make a HUGE difference - that you just have to take anything with a shaker or two of salt.

So, finally got off (well, actually on) my butt and did some testing I've been thinking about for a while, writing down all the numbers as they showed up. Yep, with pen and paper (the horror!) I should note that I'm still using CDM 3.0.1.

First, I took my CM9 card (SanDisk 16GB Ultra 1) and ran the standard 5x, 50MB, 4kQD32 test a few times. Saw the usual variations and the high score posted as the result.
Next, I ran the exact same test, except I ran the 1x option. 20 times. I won't bore you with actual numbers, but the results were ridiculously tight, and all on the higher end of the range. If I get really motivated at some point, I'll do some basic statistics and see what they say.

Now, of course, this was only testing the initial 300MB boot partition of the card, so I did the next logical thing ;) and bought a brand stinkin' new card. SanDisk 32GB class 10, UHS1, from Costco -- SDSDQU-032G-AC11A for those keeping score at home.The full test revealed, of course 10x+ serial write speed.

So, I ran the 4kQD32 50MB test again, this time set to 9 iterations. The read number stayed fairly steady, around 4.165 ish, +/- .01. Then, as I watched, the write number started high, at 2.184 (!) and steadily declined to 1.889 (!). I mean, it was a nearly linear drop. Of course, the final result was the 2.184 figure (NEVER seen one this high in my setup before, btw, even other UHS1 cards) Lather, rinse, repeat with nearly identical results, only a tiny bit higher this time, both read and write-- 4.25 and 2.213>1.891.

Next, the 1x iteration, 15 times. Read speeds centered around 4.2ish, but again, the write results were nearly consistent, around 2.2, +/- .01.

Huh....

So, I shrank the test area (first/only partition visible in Windows) from the full 32GB down to 2.8GB, to sort of mimic the "system/data" area of a CM SD install.

Darned if the results weren't virtually identical. Same exact behavior in both 1x and 9x tests.

So, what do we gather?

Well, obviously, there are some technicalities that would be nice to know about how CDM does its multiple iteration testing. Why does the number drop, is this somehow directly testing progressively across the whole memory space (doubt it), or is the wear-leveler in the chip somehow creating the discrepancy/drop? My from-the-hip-no-leg-to-stand-on guess would be that the wear-leveling algorithm/handler is the whole key to the success of the SanDisk where others fail, regardless of whether it's at the root of these results or not.

This card tests ridiculously fast, I'd love to find a meaningful way to benchmark it in use in an SD install.

I started researching for a better test, found something incredibly technical and very comprehensive which I have not yet figured out how to use, and got to wondering... what file size and Queue Depth would best emulate the usage of a Nook Color SD install? Started trying to find that out, and ran out of steam. I'm figuring there's some Android genius (not a genius android, but... you know what I meant!) out there who actually knows something about this without me breaking my brains over it (not that I don't like learning... :) )

There was some other observation I had, but I'm not recalling it atm... If I think of it, I'll edit later.
If I manage to get the time, I'll do some poking around in a Linux environment and see what I can learn. Or... anyone? Bueller?
 
  • Like
Reactions: lmacmil

Top Liked Posts

  • There are no posts matching your filters.
  • 34
    1/16/2012 update: Thanks to waxhell for compiling all of the results in this thread (at least as of mid-december! OK, so I took a while to get this posted...)
    https://docs.google.com/spreadsheet/ccc?key=0AjPE3ZAD2eVudE9vZmQ3aHlfTkFvU2J4ZUplRDJQTEE

    5/20 update: SD card performance benchmark table by a.fenderson from later in this thread added at the bottom of this post. Thank you a.fenderson!
    ------------------
    April minor update with a Transcend SD (see http://forum.xda-developers.com/showpost.php?p=12964262&postcount=8)

    _________________________________________________

    Let me tell you my story.

    Over the last couple of months as I have mucked around with SD booting various flavors of froyo and, more recently, CM7, I have found it maddening that there are apparently so many people that LOVE those versions, even people that seemed to have the same SD card as the two I've tried.

    One of those is a Class 10 Patriot 8gig, which I've seen mentioned as an acceptable boot disk.
    The other is a Kingston Class 4 4gig that tests out as significantly slower than the Patriot, but didn't really run froyo any slower.

    Now I know I've seen posts that mention small block write speed as being important, but the numbers I've seen posted really didn't make me think the Patriot was the cause...
    ...the cause of FRUSTRATINGLY slow UI, where froyo (and CM7) seem to go off into lala-land for a few seconds every time I (tried to) do something.
    ...the cause of glacial web-surfing, where even downloading GOOGLEforchrissake takes forever. Of snail's-pace market downloads. Oh god.

    Every time I would give up and go back to Eclair and breath a sigh of relief, reveling in the snappiness, the zippy web-surfing, the rapid response of m.pornhub.com.

    And every time the siren song of the CM7-elite would call me back. THIS time I did something different. I saw mention that a Class 2, YES A FREAKING CLASS 2, Sandisk from costco worked well. So I skippity-skopped up and bought one.

    After backing up the Patriot and writing THAT VERY SAME img to the Sandisk (which, yes, took 3 times as long as writing to the Patriot), I booted CM7.

    OMG

    The heavens opened and the angels sang, I HAD ARRIVED IN THE PROMISED LAND!
    NOW I knew what the chosen had been praising! NOW I knew the joy that is CM7! ZIP-ZIP-ZIPPITY!

    HOLY CRAP?! How could a benchmark be so misleading? I HATE those bastard SD manufacturers, or maybe it's the industry group that chose such a sucky measure of speed.

    For your edification, here are some CrystalDiskMark (5 iterations, 50MB) results from 2 sucky cards and the good Sandisk.


    __Test_________________Sandisk 8G Class 2___Patriot 8G Class 10____Kingston Class4 4Gig
    Sequential Read :__________10.871 MB/s________20.036 MB/s___________18.700 MB/s
    Sequential Write :__________6.659 MB/s________13.660 MB/s____________4.277 MB/s
    Random Read 4KB (QD=32) :___3.077 MB/s_________3.444 MB/s____________2.088 MB/s
    Random Write 4KB (QD=32) :__1.791 MB/s________0.038 MB/s_____________0.016 MB/s


    Look at that, both the Kingston and the Patriot blow away or are close enough to the Sandisk in everything BUT....

    (wait for it)....

    SMALL BLOCK RANDOM WRITES
    where the Sandisk is FIFTY times the speed of the Patriot and ONE HUNDREDtimes the Kingston.

    Now you may say, "Swizzlenuts, old pal, I KNEW that."

    But for all you poor slobs who didn't, who are dragging yer sorry ass through the broken glass that is booting off of a slow SD (you know who both of you are), I hope this helps.

    And now you may commence posting links to specific posts where this info was discussed in detail last December.


    ------------------
    Thanks to all of the people that posted their results and to a.fenderson for compiling them. Here's his table from http://forum.xda-developers.com/showpost.php?p=13991898&postcount=144
    Code:
    POSTER         BRAND      CLASS  CAPACITY  4 KB RANDOM
                                      (GB)    WRITE QD32(MB/s)    NOTE
    swoozle        SanDisk        2    8          1.791           Model: SDSQ-8192-AC11M
    a.fenderson    SanDisk        4    16         1.660           SanDisk C4 16GB 1 of 2 cards
    MickMcGeough   SanDisk        4    8          1.59            benchmarked in XBench, QD unknown
    a.fenderson    SanDisk        4    16         1.500           SanDisk C4 16GB 2 of 2 cards
    Awats          SanDisk        4    16         1.391           SanDisk C4 16GB 1 of 2 cards ; fastest (onboard) reader
    arwild01       SanDisk        4    8          1.369    
    Awats          SanDisk        4    16         1.284           SanDisk C4 16GB 2 of 2 cards ; fastest (onboard) reader
    robot8         SanDisk        4    8          1.270           SanDisk C4 8GB 2 of 2 cards
    robot8         SanDisk        4    8          1.259           SanDisk C4 8GB 1 of 2 cards
    chinly43       SanDisk        4    16         1.257           (not via Nook) ; same card as chinly43's other listed SanDisk C4 16GB
    joobu          SanDisk        4    4          1.175    
    angomy         Nook internal  N/A N/A         1.116           via Nook on USB
    chinly43       Nook internal  N/A N/A         1.094           via Nook on USB
    a.fenderson    Transcend      2    32         1.032    
    a.fenderson    SanDisk        4    4          0.898    
    a.fenderson    SanDisk        2    4          0.891    
    pchoi94        SanDisk        4    16         0.834    
    chinly43       SanDisk        4    16         0.769           via Nook on USB ; same as chinly43's other listed SanDisk C4 16GB
    pryonix        SanDisk        4    8          0.625    
    chinly43       SanDisk        2    8          0.616    
    a.fenderson    SanDisk        4    8          0.596    
    angomy         SanDisk        4    16         0.574    
    Blue6IX        SanDisk        2    16         0.350    
    a.fenderson    SanDisk       N/A   2          0.269    
    Awats          SanDisk        4    2          0.261           SanDisk C4 2GB 1 of 2 cards ; fastest (onboard) reader
    Awats          SanDisk        4    2          0.236           SanDisk C4 2GB 2 of 2 cards ; fastest (onboard) reader
    pryonix        Samsung       N/a   2          0.093    
    victle         Kingston      N/A   2          0.051    
    victle         Dane-Elec     N/A   2          0.050           0.050 or less:  exact value unspecified
    joobu          Lexar          4    8          0.038    
    swoozle        Patriot        10   8          0.038    
    Blue6IX        Dane-Elec      4    4          0.037           (made in Japan)
    omghahalol     Transcend      6    16         0.037    
    Ravynmagi      Samsung(??)    2    4          0.037    
    pryonix        Kingmax        10   16         0.036    
    Ravynmagi      Wintech        10   16         0.036    
    victle         Transcend      6    4          0.034    
    omghahalol     Transcend      6    8          0.033    
    Ravynmagi      SanDisk        2    8          0.033    
    pryonix        Transcend      6    8          0.033    
    Ravynmagi      Patriot        10   16         0.030    
    Blue6IX        PNY            10   8          0.030    
    pchoi94        Kingston       2    16         0.030    
    swoozle        Transcend      6    8          0.029           Model: TS8GUSDHC6
    joobu          ??            N/A   2          0.029    
    chinly43       SanDisk       N/A   1          0.029    
    a.fenderson    SanDisk       N/A   1          0.028    
    victle         PNY            4    8          0.028    
    Blue6IX        PNY           N/A   2          0.027           (made in Taiwan)
    omghahalol     SanDisk        2    2          0.021    
    MickMcGeough   SanDisk        2    8          0.02           benchmarked in XBench, QD unknown
    robot8         Transcend      6    8          0.018           Transcend C6 8GB 1 of 2 cards
    ExploreMN      Patriot        10   16         0.018    
    arwild01       Samsung(??)    4    8          0.017    
    swoozle        Kingston       4    4          0.016           Model: SDC4/8GB
    Awats          Patriot        4    4          0.016           Patriot C4 4GB 1 of 2 cards ; fastest (onboard) reader
    robot8         Transcend      6    8          0.014           Transcend C6 8GB 2 of 2 cards
    Tnexus         Patriot        10   16         0.014    
    a.fenderson    Kingston       4    8          0.014    
    Awats          Patriot        4    4          0.011           Patriot C4 4GB 2 of 2 cards ; fastest (onboard) reader
    chinly43       Lexar          4    8          0.011
    6
    This thread should be pinned: specifically the list of cards and their bench results. I was amazed at what a difference changing my SD card made.

    I fear swoozle (the OP) may have abandoned us, but if not or if an admin who wants to sticky this thread could do so, I'd be glad for someone to copy and paste the list I keep updating to the end of the first post. Speaking of that:

    Code:
    POSTER		BRAND		CLASS	CAPACITY (GB)	4 KB RANDOM WRITE QD32 (MB/s)	NOTE
    swoozle		SanDisk		2	8		1.791	
    a.fenderson	SanDisk		4	16		1.660				SanDisk C4 16GB 1 of 2 cards
    MickMcGeough	SanDisk		4	8		1.59				benchmarked in XBench, QD unknown
    a.fenderson	SanDisk		4	16		1.500				SanDisk C4 16GB 2 of 2 cards
    Awats		SanDisk		4	16		1.391				SanDisk C4 16GB 1 of 2 cards ; fastest (onboard) reader
    arwild01	SanDisk		4	8		1.369	
    Awats		SanDisk		4	16		1.284				SanDisk C4 16GB 2 of 2 cards ; fastest (onboard) reader
    robot8		SanDisk		4	8		1.270				SanDisk C4 8GB 2 of 2 cards
    robot8		SanDisk		4	8		1.259				SanDisk C4 8GB 1 of 2 cards
    chinly43	SanDisk		4	16		1.257				(not via Nook) ; same card as chinly43's other listed SanDisk C4 16GB
    joobu		SanDisk		4	4		1.175	
    angomy		Nook internal	N/A	N/A		1.116				via Nook on USB
    chinly43	Nook internal	N/A	N/A		1.094				via Nook on USB
    a.fenderson	Transcend	2	32		1.032		
    a.fenderson	SanDisk		4	4		0.898	
    a.fenderson	SanDisk		2	4		0.891	
    pchoi94		SanDisk		4	16		0.834	
    chinly43	SanDisk		4	16		0.769				via Nook on USB ; same as chinly43's other listed SanDisk C4 16GB
    pryonix		SanDisk		4	8		0.625	
    chinly43	SanDisk		2	8		0.616	
    a.fenderson	SanDisk		4	8		0.596	
    angomy		SanDisk		4	16		0.574	
    Blue6IX		SanDisk		2	16		0.350	
    a.fenderson	SanDisk		N/A	2		0.269	
    Awats		SanDisk		4	2		0.261				SanDisk C4 2GB 1 of 2 cards ; fastest (onboard) reader
    Awats		SanDisk		4	2		0.236				SanDisk C4 2GB 2 of 2 cards ; fastest (onboard) reader
    pryonix		Samsung		N/a	2		0.093	
    victle		Kingston	N/A	2		0.051	
    victle		Dane-Elec	N/A	2		0.050				0.050 or less:  exact value unspecified
    joobu		Lexar		4	8		0.038	
    swoozle		Patriot		10	8		0.038	
    Blue6IX		Dane-Elec	4	4		0.037				(made in Japan)
    omghahalol 	Transcend	6	16		0.037	
    Ravynmagi	Samsung(??)	2	4		0.037	
    pryonix		Kingmax		10	16		0.036	
    Ravynmagi	Wintech		10	16		0.036	
    victle		Transcend	6	4		0.034	
    omghahalol 	Transcend	6	8		0.033	
    Ravynmagi	SanDisk		2	8		0.033	
    pryonix		Transcend	6	8		0.033	
    Ravynmagi	Patriot		10	16		0.030	
    Blue6IX		PNY		10	8		0.030	
    pchoi94		Kingston	2	16		0.030	
    swoozle		Transcend	6	8		0.029	
    joobu		??		N/A	2		0.029	
    chinly43	SanDisk		N/A	1		0.029	
    a.fenderson	SanDisk		N/A	1		0.028	
    victle		PNY		4	8		0.028	
    Blue6IX		PNY		N/A	2		0.027				(made in Taiwan)
    omghahalol 	SanDisk		2	2		0.021	
    MickMcGeough	SanDisk		2	8		0.02				benchmarked in XBench, QD unknown
    robot8		Transcend	6	8		0.018				Transcend C6 8GB 1 of 2 cards
    ExploreMN	Patriot		10	16		0.018	
    arwild01	Samsung(??)	4	8		0.017	
    swoozle		Kingston	4	4		0.016	
    Awats		Patriot		4	4		0.016				Patriot C4 4GB 1 of 2 cards ; fastest (onboard) reader
    robot8		Transcend	6	8		0.014				Transcend C6 8GB 2 of 2 cards
    Tnexus		Patriot		10	16		0.014	
    a.fenderson	Kingston	4	8		0.014	
    Awats		Patriot		4	4		0.011				Patriot C4 4GB 2 of 2 cards ; fastest (onboard) reader
    chinly43	Lexar		4	8		0.011
    4
    Hi,

    I'm getting my nook next week, and I'm trying to pick an SD for it. I initially wanted an 8GB Class 6 Adata, but they're quite difficult to get a hold of in the UK.

    I'm torn between a SanDisk Class 4 and a Samsung Class 6. Both are 8GB.

    Anyone got any experience with Samsung or should I still to the tried and tested SanDisk?

    edit: I managed to find somewhere with a Class 6 Adata, so that's still an option.

    Any advice would be much appreciated.

    Will you be running your Android install off the SD card, or just using it for data?
    If just data, you'll be fine for most applications.

    I benchmarked several brand-new SanDisk cards of varying capacities, and all the >= 4 GB cards were Class 4, but even so it was hit-or-miss: random 4K writes for the 1 and 2 GB cards sucked, the 4 GB card was decent, the 8 GB card was half as fast as the 4 GB, one of the 16 GB cards looked great, and the other had kind of inconsistent results. So even within the same brand (and class rating where applicable), there is a lot of variation, unfortunately. You might look below and narrow your purchasing options down to those combinations of brand, size, and Class whose 4K random writes are shown to be around 1.00 MB/s or better, though really there are two orders of magnitude difference between the truly crappy cards and the best ones, so even cards with 0.1 to 0.9 MB/s 4K random writes might suffice when compared to the 0.0X cards. In general, so far it looks like SanDisk Class 4 and 2 vary between good (~.5 MB/s) and excellent (~1.5 MB/s), though unclassed SanDisk aren't so great.

    If running CM7 (etc) off it, my best advice is to test whatever card(s) you do end up buying with CrystalDiskMark and only using it as your boot disk if the 4K random writes are approximately equal to or greater than 1.00 MB/s.

    @all: the anecdotal reports of your cards running well are great, but if you have the time, please benchmark your cards with CrystalDiskMark so that we can all get enough data to be more confident in our card purchases. If you're worried about AV warnings when you download, please see this post where I explained what the opencandy component of the installer does and does not do (it's not malware). Also, along with your benchmarks, please provide your evaluation of how well the card runs your ROM, etc, if applicable.

    Here's a compilation of the actual benchmarking data thus far in this thread (with the only benchmakred data presented as 4 KB random write w/ QD=32, since this is the only commonality among all posters). It's been sorted (descending) on the write speed column:

    Code:
    POSTER		BRAND		CLASS	CAPACITY (GB)	4 KB RANDOM WRITE QD32 (MB/s)
    swoozle		SanDisk		2	8		1.791
    a.fenderson	SanDisk		4	16		1.660
    a.fenderson	SanDisk		4	16		1.500
    robot8		SanDisk		4	8		1.259
    chinly43	SanDisk		4	16		1.257
    joobu		SanDisk		4	4		1.175
    angomy		Nook internal	N/A	N/A		1.116
    a.fenderson	Transcend	2	32		1.032
    a.fenderson	SanDisk		4	4		0.898
    chinly43	SanDisk		2	8		0.616
    a.fenderson	SanDisk		4	8		0.596
    angomy		SanDisk		4	16		0.574
    a.fenderson	SanDisk		N/A	2		0.269
    swoozle		Patroit		10	8		0.038
    joobu		Lexar		4	8		0.038
    swoozle		Transcend	6	8		0.029
    joobu		??		N/A	2		0.029
    chinly43	SanDisk		N/A	1		0.029
    a.fenderson	SanDisk		N/A	1		0.028
    robot8		Transcend	6	8		0.018
    ExploreMN	Patroit		10	16		0.018
    swoozle		Kingston	4	4		0.016
    Tnexus		Patroit		10	16		0.014
    a.fenderson	Kingston	4	8		0.014
    chinly43	Lexar		4	8		0.011
    2
    Thanks a lot for clarifying, I will try it out now that I know I can opt out of the install.

    Edit: Ran CrystalDiskMark for 16GB class 4 Sandisk card vs internal nook storage via USB storage mode:

    Sandisk 16GB class 4 (non-Ultra+)

    Sequential Read : 20.003 MB/s
    Sequential Write : 3.086 MB/s
    Random Read 512KB : 20.113 MB/s
    Random Write 512KB : 1.372 MB/s
    Random Read 4KB (QD=1) : 5.143 MB/s [ 1255.7 IOPS]
    Random Write 4KB (QD=1) : 0.594 MB/s [ 145.1 IOPS]
    Random Read 4KB (QD=32) : 6.019 MB/s [ 1469.4 IOPS]
    Random Write 4KB (QD=32) : 0.574 MB/s [ 140.1 IOPS]


    Nook internal:

    Sequential Read : 19.970 MB/s
    Sequential Write : 3.937 MB/s
    Random Read 512KB : 20.634 MB/s
    Random Write 512KB : 2.173 MB/s
    Random Read 4KB (QD=1) : 4.905 MB/s [ 1197.5 IOPS]
    Random Write 4KB (QD=1) : 1.106 MB/s [ 270.0 IOPS]
    Random Read 4KB (QD=32) : 4.446 MB/s [ 1085.4 IOPS]
    Random Write 4KB (QD=32) : 1.116 MB/s [ 272.4 IOPS]
    2
    Update:Transcend

    Just a minor addition with a Transcend card. I'd seen many posts that swore by a Transcend Cl 6 8G card through Newegg. I ordered, received, benchmarked and loaded CM7 Nightly 37.
    To cut to the chase, the Transcend sucked big hairy monkey balls. Similar small block write speeds as the other crappy cards. And the CM7 performance was predictably bad. Crappy UI response and FCs.

    __Test_________Sandisk Cl2 8G___Patriot CL10 8G____Kingston Cl4 4Gig___Transcend Cl6 8G
    Sequential Read___10.871 MB/s____20.036 MB/s___________18.700 MB/s___________19.930 MB/s
    Sequential Write___6.659 MB/s____13.660 MB/s____________4.277 MB/s___________19.325 MB/s
    Rndm Rd 4KB(QD32)__3.077 MB/s_____3.444 MB/s____________2.088 MB/s___________2.968 MB/s
    Rndm Wrt 4KB(QD32)_1.791 MB/s_____0.038 MB/s____________0.016 MB/s___________0.029 MB/s

    I'm sure any of these would work fine as data cards. But for running off of SD, it makes a huge difference.