Upgrading To Faster Cf Card Options

[fquindos]

Well, no. There is not a magic limit of 4 hours uptime as I suspected yesterday late in the night :-) In fact, the last test I did gave me 6 hours! (Again almost exact, oh well). The difference with the last one was that the wifi card was not connected.

So that's what I did: I installed a crontab like this one in my user account:

* * * * * (/bin/date; /usr/bin/apm) >> /home/fquindos/up.txt

Then, with the yellow led off, I disconected the power supply and closed the lid. This morning the unit was off (surprise). Clicking the power button the green led went on for about a minute, but nothing appeared on the display. Not possible to start the unit without power supply. Examining the up.txt log file I found:

Battery state high, 100% remaining from May 31 22:57 to Jun 1 01:30.
Battery state high, 95% remaining at Jun 1 01:31.
Battery state high, 91% remaining from Jun 1 01:32 to Jun 1 01:36.
Battery state high, 100% remaining from Jun 1 01:37 to Jun 1 01:45. (Really).
Battery state high, 95% remaining from Jun 1 01:46 to Jun 1 01:48.
Battery state high, 100% remaining at Jun 1 01:49.
Battery state high, 95% remaining from Jun 1 01:50 to Jun 1 01:55.
Battery state high, 91% remaining at Jun 1 01:56.
Battery state high, 95% remaining from Jun 1 01:57 to Jun 1 01:59.
Battery state high, 91% remaining at Jun 1 02:00.
Battery state high, 95% remaining from Jun 1 02:01 to Jun 1 02:02.
Battery state high, 91% remaining from Jun 1 02:03 to Jun 1 04:51.
Battery state high, 87% remaining at Jun 1 04:52.
Battery state high, 83% remaining at Jun 1 04:53.
Battery state high, 75% remaining from Jun 1 04:54 to Jun 1 04:56.
Battery state high, 50% remaining from Jun 1 04:57 to Jun 1 04:59.
Battery state charging, 91% remaining at May 31 22:54. (Wait, I have to adjust date :-)).

iamasmith, do you test with a network card?

Really strange. Always 4 hours exactly. I'm testing battery life in my 3200 this time with no CF card attached. I'm wondering if the battery is really empty and is not possible to start the unit without powering it. Or if after powering the unit just enough to start it, battery runs for hours (4 again?). This smells to "feature" :-)

[iamasmith]

Hmm, it may be related to the granularity of the apm driver and the possibility of apmd missing 'safe' levels if it does not suspend early.

Read apm(4) on the Zaurus. It mentions that the apm driver may be 25% off, sounds like Uwe didn't implement high granulrity on the Battery checks (might take a look at this). If this is indeed the case it may be possible that apm is suspending the unit at 25% rather than at a lower value.

-Andy

[iamasmith]

starting to play around now with apmd auto performance scaling to see if the 4 hour limit is effected by that.

and yes I'm using a Socket 10/100 Low Power Ethernet card. I wanted the disk activity to be 'background' and not contribute to the writes so I opted for a rudimentary test of running 'top' over an ssh connection and measuring uptime by seeing when it stalled.

-Andy

[iamasmith]

I started to consider the Bonnie++ test a little more... look at them again.

CODE

Bonnie++ disk performance test

Microdrive:

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
mungo.chili.lo 300M  2223  31  2297  11  1221  18  1939  93  2256  71  31.6   3
                   ------Sequential Create------ --------Random Create--------
                   -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
             files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                16  1252  48  8291 100  3263  51  1279  45  8345  99  1622  26

SanDisk 4Gb Extreme III:

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
mungo.chili.lo 300M  3162  98  3902  91  1498  95  1905  97  2414  99 100.5  17
                   ------Sequential Create------ --------Random Create--------
                   -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
             files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                16  1329  43  9492  99  4118  64  1499  53  8321  95  2869  48

Note that in the tests on the squential read we are seeing high CPU utilisation - for the SanDisk card there is a higher level of CPU utilisation for a number of operations. A trend that indicated the performance bottleneck is becoming CPU bound rather than IO bound.

So I cranked up the Zaurus speed to 520Mhz and repeated the test..

CODE

Bonnie++ disk performance test

Microdrive with Zaurus at 416Mhz (default):

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
mungo.chili.lo 300M  2223  31  2297  11  1221  18  1939  93  2256  71  31.6   3
                   ------Sequential Create------ --------Random Create--------
                   -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
             files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                16  1252  48  8291 100  3263  51  1279  45  8345  99  1622  26

SanDisk 4Gb Extreme III with Zaurus at 520Mhz:
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
mungo.chili.lo 300M  3294  97  3960  96  1550  98  2059  97  2468  99 103.3  13
                   ------Sequential Create------ --------Random Create--------
                   -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
             files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                16  1630  53 10054 100  4352  68  1824  58  9339  98  3111  45

Interestingly the CPU utilisation didn't go down.. this shows I think that the card is certainly capable of even higher transfers than the processor can handle but look at the difference in performance levels now.

startxfce4 now takes 53 seconds instead of the original Zaurus 1min 20s

-Andy

[fquindos]

...

So I cranked up the Zaurus speed to 520Mhz and repeated the test..

...

How did you this? Is overclocking the Zaurus a good idea?

[iamasmith]

...

So I cranked up the Zaurus speed to 520Mhz and repeated the test..

...

How did you this? Is overclocking the Zaurus a good idea?

Well it's not 'recommended' just as you would indemify yourself against recommending any slight gamble.

On OpenBSD check the sysctl.conf file and have a look for machdep.maxspeed this sets the upper speed. yes you can go to 624Mhz but it gets kinda hot running at that speed.

Note though that there is no physical difference in the PXA270 fitted to a 416Mhz device than the one fitted to a 624Mhz device - the processor can be safely clocked to these speeds - however, the manufacturer has several design factors for the device when they choose to clock at a specific speed. The issue is heat, power and clocking of other devices to meet the speed of the processor.

I'm actually running the device in performance scaling mode which means that I added -A to the startup flags of apmd in rc.conf. This slows the device when running on battery power if the CPU utilisation is low.

I will repeat the battery life test with this setting when the battery is fully charged again.

Check out apmd(8) for more detail on scaling.

-Andy

[iamasmith]

After the battery test I will attempt to provoke early failure (as mentioned TriSoft had a card fail after about a week). I figure that if I compile KDE on the device it will probably take about a week and provide a meaningful soak test.

-Andy

[iamasmith]

4 hours again with CPU scaling enabled.

ok, now to build KDE

-Andy

[iamasmith]

a few days into kde build now. The unit has built kdelibs and is now working on kdebase.

I have been running the unit at a constant 520Mhz with 512Mb swap (half on a partition and half on a file in the filesystem since I thought it needed a little more for this build).

The unit actually seems to be running cooler at 520Mhz with this card fitted than it ever did with the Microdrive which is encouraging.

No drive errors are present at the moment with the single exception of suspending/resuming, this still produces the 'operation failed, retrying' message that you get because suspend was instant and the rewrite had possibly been attempted before the card was ready. - This is the same as with the Microdrive so I'm not concerned at all about it.

-Andy

[albertr]

4 hours again with CPU scaling enabled.

ok, now to build KDE

-Andy

Interesting... Doesn't it look like CPU scaling not working correctly on OBSD or maybe it doesn't scale VCORE voltage? I'm running powernowd under linux 2.4.20 on the Z, and downscaling cpu freq and vcore voltage seems to greatly help to prolong battery life.
-albertr

[iamasmith]

Kde 3.5 running on OpenBSD on the Zaurus...

 kdezaur.jpg ( 36.77K ) Number of downloads: 214


I built (basically to soak test the CF card), kdelibs, kdebase and i18n-en_GB. It does actually work and it seems faster than I imagined that it would be even though it is almost constantly 20Mb overcommitted.

My feeling is that it isn't really viable to run KDE without a hardware mod on the memory but I thought I would post the picture for you to see.

-Andy

[iamasmith]

Just a quick update... I'm 10 days into running with the new card, the Zaurus has been almost constantly on and almost all of this time it has been running constantly at 520mhz - it actually seems cooler running at 520Mhz with the new card than it did with the Microdrive at 416Mhz - I think I will keep it that way.

No errors... the thing seems bullet proof.. but I will report back if I have a failure.

-Andy

[coreilly]

any news on the CF card in the Zaurus Andy???
still going good?

Just a quick update... I'm 10 days into running with the new card, the Zaurus has been almost constantly on and almost all of this time it has been running constantly at 520mhz - it actually seems cooler running at 520Mhz with the new card than it did with the Microdrive at 416Mhz - I think I will keep it that way.

No errors... the thing seems bullet proof.. but I will report back if I have a failure.

-Andy

[ins0mniaque]

I just ordered a Transcend CF 8GB 120x (they are incredibly cheap), and I intend to replace my MicroDrive with this. I'll post results as soon as I get the card...


EDIT:
I just received the card! And it's already in the Zaurus! Had trouble with the internal CF plastic cover screws, had to disconnect all ribbon cables to be able to apply pressure, and they are hell to reconnect ( at least with my big hands ).

I am installing OpenBSD 4.0-current on it right now... probably should have copied the OS from the MicroDrive, but a friend asked me how hard it was to install OpenBSD, perfect timing

Everything seems to work right now, and it seems like the same speed. The biggest noticable difference right now (apart from the 4 additional GB of storage space) is that you don't hear/feel the HD turning...

I will do some more tests this weekend...

[ins0mniaque]

Just an additional note, when I opened my Zaurus (SL-C3000), it had the copper foil strap, but it was already unsoldered and disconnected (and no, it was not my fault)! And since the only soldering iron I had when I swaped the card was a ColdHeat Pro ( I was at work, and this iron, excuse my language, sucks balls... ), so I didn't even bother to resolder it.

Right now, it's been running non-stop @ 520 MHz, mounted with noatime and softdep, building ports for over 40 hours, and everything is still running fine...