In comp.sys.ibm.pc.hardware.chips Van Chocstraw said:
You can test it by removing one strip at a time and trying
it if you are having problems.
This only works if:
1) The problem occurs often / quickly enough; and
2) You are sure that memory is the cause.
Also, absence of proof isn't proof of absence.
Many reliability problems don't occur often, or at least don't
produce visible consequences frequently. Testing is almost alwyas
more intense than regular usage and monitors continually for errors.
The cause is similarly uncertain. When you have a crash, how
can you possibly know what caused it? If it reproduces well,
then it's probably software. But if not, there still are software
interactions, and all the different peices of equipment (HD, PSU,
....) that could cause it. Testing eliminates them one-by-one.
-- Robert