On Wednesday 26 September 2012 13:14:07 Dr. Michael J. Chudobiak wrote:
> On 09/20/2012 04:33 PM, Koen Kooi wrote:
> > Op 20 sep. 2012, om 22:25 heeft Bill Mar <
bil...@specialcomp.com>
het volgende geschreven:
> >> Special Computing has known-good microSD cards tested with the
> >> BeagleBone
> >
> > That lists the exact same transcend card that I mentioned :)
>
> Transcend SD cards do seem to be the best in my tests. Here are some
> benchmarks of readily available cards:
>
>
> [root@xena flashbench]# ./flashbench -a /dev/sde
>
> apacer AP-MSD04GCS4P-TM Class 10 industrial:
> align 1073741824 pre 2.91ms on 3.6ms post 2.67ms diff
> 812盜
>
> Kingston SDC10/32GBSP:
> align 1073741824 pre 2.12ms on 2.37ms post 2.12ms diff
> 253盜
>
> Kingston SDC4/4GB:
> align 1073741824 pre 1.89ms on 2.15ms post 1.94ms diff
> 237盜
>
> SanDisk Ultra 16 GB MicroSDHC Class 10 UHS-1, SDSDQUA-016G-U46A
> align 1073741824 pre 1.46ms on 1.6ms post 1.39ms diff
> 178盜
>
> Transcend 16 GB Class 10 microSDHC Flash Memory Card TS16GUSDHC10
> align 1073741824 pre 1.42ms on 1.57ms post 1.31ms diff
> 198盜
>
> Transcend 8 GB Class 10 microSDHC Flash Memory Card TS8GUSDHC10
> align 1073741824 pre 1.37ms on 1.58ms post 1.37ms diff
> 214盜
Can you run the full suite of tests to find the eraseblock size, open-au
linear, and open-au random? And then mail the output to the flashbench-
results mailing list for each of those cards?
If you look at some existing mails to that list, it should give you a
good overview of how to determine the values that matter. The simple -a
test only does a read timing test to help find the eraseblock size, it
doesn't show anything about the write performance.
The flashbench-results list is very friendly. Arnd's really helpful and
sometimes even I can know what I'm talking about :)
As an example of a flashbench mail I sent earlier this year, see [1].
[1]:
http://lists.linaro.org/pipermail/flashbench-results/2012-August/000322.html
-Andrew