nurmi said:
For some reason the replies to this question do not appear although I
clicked on "Reply to the Group"
I have executed your suggestions and found the PC clean, the condition of
the capacitors around the processor socket appear to be in good condition,
and the memory tests OK as well. The problem continues at least once
shortly after booting up. After that it runs for hours without a problem.
I have one problem that you can fix: How to I stopp a chkdsk command? I
have searched the help programs without success.
Thanks
nurmi
Based on your symptoms right now, the next thing I'd try is swapping in
a spare ATX power supply, and retest. If this is a power supply related
issue, that might make a difference.
Testing power supplies, is too difficult to do effectively at home
(using only a multimeter). If your existing power supply had bad capacitors,
there might be short events of out of spec DC levels, just after the
power supply is started (that is what I experienced here with an Antec
brand supply). A multimeter is not the best instrument for detecting
things such as spikes on the DC rails, and other instruments are too
expensive, even to rent. You can buy a new power supply, for less than
a lot of other testing options.
*******
If you use Speedfan from almico.com , is the CPU temperature below 65C ?
http://www.almico.com/speedfan441.exe
For a stress test, you can try Prime95 from mersenne.org/freesoft . Their
site is not operating properly right now, so you'll have to wait until
later to get the file. The stress test does a mathematical calculation
with a known answer, and the program can detect if an error has been
made. This test increases the CPU temperature, and accelerates stability
testing. The test should run error free for hours - I accept a four
hour run, error free, as proof the computer is operating well. On an
unstable system, that test will stop in a couple seconds with an error.
Some stability problems, only happen when the CPU is idle. They could be
due to changing P-state or C-state on the processor, and operating it
at some combination of low frequency and low voltage (the CPU is tested
at the factory, to be stable over those conditions, so it is supposed
to work, and they include sufficient margin for aging). You could try
using the "Power Options" control panel and selected a power
scheme such as "Always On", in an attempt to keep the CPU running
in the highest state all the time. But I wouldn't bet on that fixing
it, or making any difference.
I've spent days and days in hardware labs, using logic analyzers on
computers we were building, to fix problems like this. It's very difficult
to find a root cause (unless the computer logs an event somewhere - many
failures are silent, because the computer dies before it can log anything).
Your chances of figuring it out with functional tests, are pretty
limited.
As an alternative test, you can boot a Linux LiveCD, such as Ubuntu from
ubuntu.com . It runs directly from the CD, and doesn't have to install
any software on the hard drive in order to work. The purpose of doing
this test, is to see if your OS is part of the problem or not. If
Linux remains running for hours (while you're web surfing with the
built-in browser), then perhaps some aspect of the OS is involved
after all (a WinXP problem, like a driver problem perhaps).
Good luck,
Paul
