Re: IBM microdrive.

From: David Given <dg_at_tao-group.com>
Date: Tue, 19 Feb 2002 14:42:42 +0000

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Monday 18 February 2002 19:29, you wrote:
[...]
> > Timing buffered disk reads: 64 MB in 152.29 seconds =430.34 kB/sec
> > Timing buffered disk reads: 64 MB in 199.30 seconds =328.83 kB/sec
> >
> > Timing buffer-cache reads: 128 MB in 4.46 seconds = 28.70 MB/sec
> > Timing buffer-cache reads: 128 MB in 4.15 seconds = 30.84 MB/sec
> >
> > It's slower than 10Mb ethernet! I should have stuck to NFS! It's reported
> > as a TOSHIBA MK5002MPL. Any suggestions on how to speed it up? I tried
> > turning on multcount, but that didn't have any effect, as you found out.
>
> Ouch! And I thought my card was a slug! So we have a puzzle on our
> hands. According to hdparm -v, the only thing I have turned on is
> readahead=8. Does that match your setup? Anybody else with one of
> these drives noticing bad performance?

I tried some more benchmarks.

 Timing buffered disk reads: 64 MB in 347.34 seconds =188.68 kB/sec
 Timing buffered disk reads: 64 MB in 340.25 seconds =192.61 kB/sec
 Timing buffered disk reads: 64 MB in 343.53 seconds =190.77 kB/sec

I think there's actually something *wrong* here. hdparm -v says:

/dev/hda:
 multcount = 0 (off)
 I/O support = 0 (default 16-bit)
 unmaskirq = 0 (off)
 using_dma = 0 (off)
 keepsettings = 0 (off)
 nowerr = 0 (off)
 readonly = 0 (off)
 readahead = 8 (on)
 geometry = 10390/15/63, sectors = 9818550, start = 0

[...]
> > That's the problem. I tried a 100Mb card we had here, and the iPAQ didn't
> > even notice I'd inserted it. The card was CardBus only. I'd rather not
> > order one and find out later that it didn't work.
>
> Was the card you tried listed in the compatibility list? If so, some
> feedback to whoever maintains the list might be in order.

No, it was just one we had lying around that I tried on the off-chance.
Actually it seems that the maximum throughput on the PCMCIA bus wouldn't make
it worthwhile. (Although your hdparm readings seem to indicate a higher
transfer rate than our calculations suggest is possible, so it could well be
we've miscalculated.)

Not being able to swap is a real pain, and renders the device pretty unusable
for our purposes. Is it possible to use a stock kernel on the iPAQ? I'd like
to standardise on 2.4.17, which is what we use on all the other Linux
platforms we deal with. I'm not terribly happy about using heavily patched
kernels.

- --
David Given
dg_at_tao-group.com

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE8cmRlf9E0noFvlzgRAptNAJ9PvjvXIhHcacl8kNiJQb7rSSwKxgCdH6J4
83Mxff4h0e81X1/Hv6ZJvYc=
=cr+r
-----END PGP SIGNATURE-----
Received on Tue Feb 19 2002 - 06:43:02 EST

This archive was generated by hypermail 2.2.0 : Mon Jul 25 2005 - 17:21:12 EDT