Debugging severe performance problems



Hi all!

One of my systems has a really severe performance problem. It has an AMD
Athlon XP 1700+ CPU. It has 768 MB RAM, and only very rarely swaps
anything. It has a 3Dfx Voodoo Banshee (rev 03) PCI card with 16 MB
on-card RAM. The performance problems has been there since the summer.
For some time, it ran woody with KDE 3.1 on top, then I had it tracking
sarge in April. It performed well at that point.

I've tried many different things, but it all boils down to that I think
the hardware is OK, but that I really have no clue where to start.

It is my father's system, so I'm not bothered with it too often, but the
performance is insanely bad. When it does updatedb, the system pretty
much freezes for the time it takes to scan the disk. Firefox takes more
than a minute to start, and a regular apt-get upgrade with trivial
security fixes makes the load go to above 5. My own system is similar
to this system, but has a somewhat slower CPU, and performs well. It is
definitely not because it doesn't have the hardware.

Not to get hung up in this, but I think I first saw the performance
problems as xfree86 4.3.0.dfsg.1-13 went into sarge. I have therefore
been working under the hypothesis that it is connected to some X
problems, and it seems like it just takes a long time to draw anything.
It is hard to tell, of course, if that's a symptom or a cause. So, I
tried replacing the 3Dfx PCI card with my own well-performing Matrox
G450 AGP card, but that had no appreciable effect. Reducing the
resolution to 800x600 did however help a bit. Now, I think X might not
have anything to do with it. I have also tried a full memtest86 scan,
but no problems were spotted by it.

I've been looking through all the logs I can think of, but seen nothing
that meets the eye.

One thing that has been haunting me is this error message:
ldconfig: Cannot mmap file /lib/libslang.so.1-UFT8.
I suppose that came in with a typo (UFT8 vs. UTF8), since I forgot to
change my sources.list to stable shortly after woody was released back
in the day. However, I've lived with this since then, and it had never
any performance input before. It's quite funny that it remains though,
I've upgraded the whole system since then...

So, well, I have a system here that borders on unusable and no idea how
to fix it. Where would you start if you had a *really* slow system?

Best,

Kjetil
--
Kjetil Kjernsmo
Programmer / Astrophysicist / Ski-orienteer / Orienteer / Mountaineer
kjetil@xxxxxxxxxxxx
Homepage: http://www.kjetil.kjernsmo.net/ OpenPGP KeyID: 6A6A0BBC


--
To UNSUBSCRIBE, email to debian-user-REQUEST@xxxxxxxxxxxxxxxx
with a subject of "unsubscribe". Trouble? Contact listmaster@xxxxxxxxxxxxxxxx



Relevant Pages

  • Re: newbie: redhat 9.0 very slow
    ... Mozilla does take a long time to load ... CPU usage: Pop up a terminal and run "top". ... Having all your RAM in use is normal and desirable. ... unsubscribe mailto:redhat-list-request@redhat.com?subject=unsubscribe ...
    (RedHat)
  • Re: dumb question about upgrading the kernel and SATA disks....
    ... AMD 64 Sempron CPU and a decent hard drive and other gear etc.... ... 512 MB of RAM. ... an amd64 system but my 939 socket motherboard won't take more than 4GB. ... To UNSUBSCRIBE, email to debian-user-REQUEST@xxxxxxxxxxxxxxxx ...
    (Debian-User)
  • Re: CPU Speed
    ... I just looked in the BIOS under the Advanced section and under the CPU Speed it shows: ... Cache RAM: 512 KB ... To UNSUBSCRIBE, email to debian-user-REQUEST@xxxxxxxxxxxxxxxx with a subject of "unsubscribe". ...
    (Debian-User)
  • Server recommendations?
    ... quad core 64-bit CPU at>2.5 GHz and more than 16 GB of RAM. ... To UNSUBSCRIBE, email to debian-user-REQUEST@xxxxxxxxxxxxxxxx ...
    (Debian-User)
  • Re: Would a 2nd processor really be a waste of time???? help
    ... > buy another processor and more ram, however this is what one supplier ... > Your server has only 1 processor. ... installation including changing the bios settings and installing the smp ... The second cpu will probably help ...somewhat. ...
    (comp.unix.sco.misc)