Kingmax 60X SD card = fast swapfile

[zmiq2]

uhm...

anyone can translate the above report into something for non-sd-and-fs-experts ?  Is then the kingmax x60 1gb sd worth ?

thanks,

[kopsis]

anyone can translate the above report into something for non-sd-and-fs-experts ?  Is then the kingmax x60 1gb sd worth ?

For the 1GB Kingmax 60x SD card:

• block write speed: 1530 KB/sec at 41% CPU load
• block read speed: 833 KB/sec at 19% CPU load

This is nowhere near the card's 60x theoretical max of 9000 KB/sec and supports my conclusion that SD card performance on the Zaurus is bound not by the performance of the SD card but by the SD interface on the Zaurus itself.

Results for a run-of-the-mill PNY 512 MB SD card (on an almost identical Zaurus):

• block write speed: 787 KB/sec at 22% CPU load
• block read speed: 1341 KB/sec at 66% CPU load

Those results seem to indicate that the Kingmax may actually sacrifice read performance for improved write performance. That's great for cameras that do mostly write operations ... not so great for a computer filesystem that sees mostly read operations. It also validates the post that started this thread which was that using a swapfile on the Kingmax seemed to give good performance. A swapfile typically sees an equal number of reads and writes. The average throughput for the Kingmax is 1182 KB/sec and 30% load. For my PNY its 1064 KB/sec and 44% load. Clearly the Kingmax will perform better as swap space.

Bottom line (to me) is that if you're buying an SD card exclusively for the Zaurus, don't pay a premium for "high-speed" SD cards unless you have an "atypical" usage pattern (swap file or writing lots of data). In general, you won't see much benefit over their "normal-speed" siblings and you may actually do worse where it typically counts (read performance). If you're going to use the card in multiple devices (Zaurus & PC, Zaurus & camera) then the high speed might be worth a few extra $$.

[obscurite]

The Kingmax is one of the cheapest 1GB cards, so no extra $$ is involved in this case. I just payed $70 (plus s/h,taxes) for one. Too bad about the read speed though.



-Daniel

[bob-london]

Here's mine does anyone understand it ??

$ ./bonnie -d /mnt/card/Tunes -f -b -n 0
Writing intelligently...done
Rewriting...done
Reading intelligently...done
start 'em...done...done...done...
Version  1.03       ------Sequential Output------ --Sequential Input- --Random-
                    -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
zaurus         300M            1173   7   533   3             780
 17  30.6  91
zaurus,300M,,,1173,7,533,3,,,780,17,30.6,91,,,,,,,,,,,,,
$

[bob-london]

I think the C700 has much lower "ram" than the 760 & 860  so my experience of using the Kingmax with a 64mb swap probably gets a double benefit i.e. double the write speed, double the swap ( I also only previously had a 32mb swap)

[kopsis]

Here's mine does anyone understand it ??

$ ./bonnie -d /mnt/card/Tunes -f -b -n 0
Writing intelligently...done
Rewriting...done
Reading intelligently...done
start 'em...done...done...done...
Version  1.03       ------Sequential Output------ --Sequential Input- --Random-
                    -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
zaurus         300M            1173   7   533   3             780
 17  30.6  91
zaurus,300M,,,1173,7,533,3,,,780,17,30.6,91,,,,,,,,,,,,,
$

The trick to reading the bonnie++ output is to ignore everything except the last line of comma seperated fields. The magic decoder ring for that last line is:

• Field 1 = host name (zaurus)
• Field 5 = block write KB/sec
• Field 6 = block write CPU %
• Field 11 = block read KB/sec
• Field 12 = block read CPU %

So your results are block writes at 1173 KB/sec - 7% CPU load, block reads 780 KB/sec - 17% CPU load.

[xarope]

Here's my results from a panasonic 1GB ext2:

$ ./bonnie -d /mnt/card/ -f -b -n 0
Writing intelligently...done
Rewriting...done
Reading intelligently...done
start 'em...done...done...done...
Version  1.03       ------Sequential Output------ --Sequential Input- --Random-
                    -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
zaurus         300M            1290   6   504  61             749  57  20.4  95
zaurus,300M,,,1290,6,504,61,,,749,57,20.4,95,,,,,,,,,,,,,

So I guess that makes it 1290kbs@6% CPU write, 749kbs@57% CPU read?

Reads seems too high in CPU, I'll probably have to rerun it again, my screen saver did pop up a few times (didn't realise the benchmark would take SO long to run!).

[edit]
Here's the rerun:

$ ./bonnie -d /mnt/card/ -f -b -n 0
Writing intelligently...done
Rewriting...done
Reading intelligently...done
start 'em...done...done...done...
Version  1.03       ------Sequential Output------ --Sequential Input- --Random-
                    -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
zaurus         300M            1338   70   505  9             750  18  20.4  95
zaurus,300M,,,1338,70,505,9,,,750,18,20.4,95,,,,,,,,,,,,,

Now it says writes take 70% CPU time?!?  yet another run time...

Version  1.03       ------Sequential Output------ --Sequential Input- --Random-
                    -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
zaurus         300M            1318   24   505  9             759  18  20.6  96
zaurus,300M,,,1318,24,505,9,,,759,18,20.6,96,,,,,,,,,,,,,

ok this seems better?!?

[/edit]

[suid]

Just got a Transcend 1Gb 60x SD and now feeling rather foolish after reading that  an Ultra-type card isn't going to give me much of an I/O improvement.

Anyway, here's the results from bonnie :

zaurus, 300M,,, 1378, 23, 838, 10,,, 1531, 24, 43.8, 92,,,,,,,,,,,,,

That's out of the box, fat formatted. I can't vouch for its reliability or how fast it actually feels yet, because I've only just got it.