How does one diagnose a lock-up?

[Wiley Q. Hacker]

I've got an older P4 1.7 GHz machine. Recently, I've been having some
trouble, where the machine just locks up. Can't move the mouse,
keyboard won't work, the off-switch won't turn the computer off.
Totally unresponsive. The only thing that works is to hold the power
switch down for 5 seconds to do a hard shutdown.

The lock-up is semi-predictable. It seems to happen after about 5-15
minutes of heavy CPU activity. Most recently, it died several times
trying to transcode some slideshows into NTSC video using Nero, Roxio
and Picasa (so it's not software-specific). It has also behaved like
this when doing full system scans using Norton AntiVirus.

Most mornings, I can turn the monitor on to find the system in this
state. This probably due to the fact that my spyware and virus scans
try to run overnight.

I installed SpeedFan during my tasks to try and see if the CPU is
overheating (since heavy CPU activity causes the issue), but the
motherboard does not seem to be reporting on CPU temperature; probably
too old.

Any pointers on diagnosing the problem?

 

[JAD]

I've got an older P4 1.7 GHz machine. Recently, I've been having some
trouble, where the machine just locks up. Can't move the mouse,
keyboard won't work, the off-switch won't turn the computer off.
Totally unresponsive. The only thing that works is to hold the power
switch down for 5 seconds to do a hard shutdown.

The lock-up is semi-predictable. It seems to happen after about 5-15
minutes of heavy CPU activity. Most recently, it died several times
trying to transcode some slideshows into NTSC video using Nero, Roxio
and Picasa (so it's not software-specific). It has also behaved like
this when doing full system scans using Norton AntiVirus.

sounds heat related or memory
what OS?

 

[johns]

When I get something like this, the first thing I do is
mope mouth it to the customer. Reason, I'm never going
to exactly pin it down, and by now the thing is an out-
of-date piece of crap, and he needs to quit being a
cheap-skate and get into a better box. If there is nothing
I can do to convince him, then at least I can get away
with sacking the thing and doing a complete reinstall.
I start with running chkdsk c: /f to see if the hard drive
is bad. Also, if I can hear the bearing screaming ... or
if the stupid thing is a 10 gig ( or 20 gig for that matter ),
it is new hard drive, "or else!" If the psupply fan is
noisy and buzzing on startup ... new psupply, "or else".
If the box is loaded with dust, I take a leaf blower to it.
If he can't "find" his OS install disk, tough! Buy one.
If his AV is out of date. Buy one. If the floppy won't
read. Buy one. If the cdrom won't read, or gives constant
errors during the install. Buy one. If he has the usual
128 meg of ram, I up it to 500 meg ... for a fee. If his
video card is an 8 meg, that goes too. By the time I
get through with this thing, it will be running fine, and
he will have bought a new computer.

johns

 

[The Seabat]

Remind me not to take my box to you for service!

 

[jcase]

In other words, you have the "customer" pay for you unwillingness or
inablility to diagnose & correct the problem.
I have several people who don't want to spend money on a new pc or any
upgrades. I make them
understand the limitations of the hardware but the pc can still be made
to work right.

 

[David Maynard]

When I get something like this, the first thing I do is
mope mouth it to the customer. Reason, I'm never going
to exactly pin it down, and by now the thing is an out-
of-date piece of crap, and he needs to quit being a
cheap-skate and get into a better box. If there is nothing
I can do to convince him, then at least I can get away
with sacking the thing and doing a complete reinstall.
I start with running chkdsk c: /f to see if the hard drive
is bad. Also, if I can hear the bearing screaming ... or
if the stupid thing is a 10 gig ( or 20 gig for that matter ),
it is new hard drive, "or else!" If the psupply fan is
noisy and buzzing on startup ... new psupply, "or else".
If the box is loaded with dust, I take a leaf blower to it.
If he can't "find" his OS install disk, tough! Buy one.
If his AV is out of date. Buy one. If the floppy won't
read. Buy one. If the cdrom won't read, or gives constant
errors during the install. Buy one. If he has the usual
128 meg of ram, I up it to 500 meg ... for a fee. If his
video card is an 8 meg, that goes too. By the time I
get through with this thing, it will be running fine, and
he will have bought a new computer.

johns

Thanks for the warning.

 

[Wiley Q. Hacker]

Thanks for all the pointers. Here's my POA based on the feedback.

1. Not being able to measure the CPU temperature, I'll just open up the
case, point a desk fan at it, and let rip my movie transcoding software.
Let's see if we get a lock-up.
2. Pull out one RAM module at a time (I've got 3 256 MB RDRAM modules), and
do the same. See what happens.

Hopefully, something will click.

 

[Tweek]

If you are running Windows XP it is a little easier because it isn't the
type of OS that will lock up just because you looked at it funny. The first
thing I do is open the case and check for bad caps on the mainboard(also
check for excessive dust clinging to the heatsink). If that looks ok, I boot
the Ultimate Boot CD and run memtest to make sure the memory is ok Then I
run Hitatchi's Drive Fitness Test to make sure the HDD is ok. If that checks
out I boot to a custom Windows Ultimate Boot CD (basically Bart PE with a
bunch of goodies). I run a chkdsk to make sure the file system is ok, then a
virus scan to eliminate that possibility. The cool thing about the WUBCD is
that it is a bootable version of XP. You can run Prime95 and a couple of
other torture tests. If you work the system over good and don't get a lockup
then generally you can eliminate a hardware problem from the equation. If I
do get a lockup and the system is clean inside and doesn't seem to be
overheating my first suspect is the power supply and I will plug in a good
one and re-run the torture tests. From there it is a process of elimination
and can take awhile. If it doesn't lockup with the WUBCD, I will boot the
machine into safe mode and check for spyware and try to reproduce the
problem there. If no lockups then it is on to normal mode to find out what
is causing it. If there is nothing obvious I usually let the customer know
that a format and restore may be the best option rather than spending hours
tracking it down. Personally I don't like to give up and enjoy a challenge,
but sometimes it is not worth the time and headache. Anyway, you can get the
UBCD here at www.ultimatebootcd.com
There is a link there to the windows ultimate boot cd with instructions how
to build it.

 

[Wiley Q. Hacker]

This really helped, Tweek.

MemTest 86, one of the tools on the UBCD, locked up in the middle of the
operation. This suggests a memory problem, doesn't it?

Problem is, I've got 3 RDRAM DIMMs. How do I narrow it down to one?

 

[Tweek]

Rdram has to be installed in matched pairs, are you sure you have three? If
you only have two, I am not aware of any way to test them indivdually.
However, a lockup in memtest doesn't necessarily indicate bad memory, it
still could be something else hardware wise. Unfortunately the only way to
eliminate memory as the culprit is to put known good modules in the machine
and run memtest again. You can also try reseating the memory and running
memtest again.