Mervyn said:
I am getting, several times a week, situations when the PC totally freezes
with no input via the mouse or keyboard but with the screen just as it was
before the freeze. Powering down fixes it.
Can anyone advise how to go about identifying and fixing the problem?
There is one other test that remains to be done.
It is possible, for a computer to have a "GUI failure". That
means the display, keyboard, and mouse could suddenly be ignored
by the OS. If all of that, funnels through one piece of software
in the OS, it means the screen might stop updating, if that
piece of software dies for some reason.
If you have a second computer, and you know the range of IP
addresses being used by your computers, you can use the
second computer to do a "ping test". That sends a test packet
to the frozen computer, and it would send a packet back
in response. It's part of the ICMP protocol on the Internet.
Say, for example, my router has DHCP, and serves local unrouted
addresses when the computers boot, in the 192.168.100.1 to 192.168.100.4
range. The second computer happens to be 192.168.100.3 at the
moment. (Go to command prompt, and type "ipconfig", to determine
the current value of the IP address of the second computer.)
You might not know, what the IP address of the frozen computer
is, but if DHCP is range limited to a few IP addresses, you
might guess at the address being used. Then, with the second
computer, try pinging the frozen one. I'm using three
separate tests here, to try and find the frozen computer.
If your computer had a static address which you knew, then
you'd only need one test line like this.
ping 192.168.100.1
ping 102.168.100.2
ping 102.168.100.4
If you get a response from the first computer, at one of those
addresses, it tells you just the GUI is frozen, and the OS
is not completely dead.
The distinction is important, in terms of your debugging process.
If the frozen computer did not respond to any ping command, then
it is likely a CPU freeze. If the frozen computer was able to
respond to an Ethernet ping, then it could be the video card
is the problem. Possible solutions might be uninstalling the
video driver, and installing another.
You can also use differential analysis, to shed light on the problem.
Boot another OS (a Linux LiveCD), and see whether the computer will
freeze there or not. If the computer doesn't freeze in Linux, but
does freeze in Windows, it could be the Windows OS install which
is damaged in some way.
I had a computer once, that froze in both Windows and in Linux,
and based on that, I could conclude it was a hardware problem.
(As it turns out, an actual design flaw in the motherboard.)
You can run memtest86+ to test the memory. You can run Prime95
(mersenne.org/freesoft) in either Windows or Linux, as a stress
test that checks correctness of computation. That tests CPU,
Northbridge, and memory.
As some of the other respondents have pointed out, instability
can be caused by bad capacitors on the motherboard, in the
Vcore area. Certain recent Dell computers suffered from that,
and other brands to different extents. Bad capacitors can
also exist inside the power supply. (I have a second Antec
power supply, that appears to be failing - I haven't opened it
up yet, but I can smell something.) So it could be a motherboard
problem, or a power supply problem. CPUs are generally pretty
good, in terms of not failing while in service. RAM less
so (I've had lots of RAM go bad here, relatively speaking,
compared to CPUs with zero failures - the RAM fails while
in service, which is ridiculous).
Depending on your funding, you can rush out and start
replacing things. Or you can spend a bit more time
testing, to isolate to what you think is the area of
the failure. It all depends on how much money you
have to throw around, as to what you do next.
There is no guarantee, that an OS log file has an entry in
it, corresponding to the time of failure. If you were
getting a real CPU freeze for some reason, the OS doesn't
always have an opportunity to record an event. So you
may not have anything to go on, in that respect. So while
you could check Event Viewer, look for memory dump files or
the like, that's not likely to be an option for you. If it
was just a GUI failure, then some event might be recorded
(where the OS can't do something to the display etc).
You may also want to run CHKDSK, to verify the file system
is OK. You can run CHKDSK, just to see whether there are
any structural errors. That can happen from all the
freezing, without cleanly shutting down the file system.
And also make sure your backups, are up to date.
Just in case.
Paul
