Before fixing things, first keep doing what you were doing.
First identify the problem. For example, you ran
comprehensive hardware diagnostics to find an intermittent.
Good. But many intermittents even pass diagnostics when
running in 70 degree F rooms. Effective testing is called
burn-in testing. Run those diagnostics in a 100 degree room
or run diagnostics while heating the tested component with a
air dryer on high. Hot enough to be uncomfortable to touch
but not leave skin is a normal operating temperature.
Intermittents become more obvious when component operates at
the extreme limits of normal temperatures.
Especially necessary to find intermittent SIMM memory is
heat. Either run Memtst86 or the manufacturer's comprehensive
memory diagnostic - at elevated, normal temperature - to
verify memory integrity. Again intermittent memory can pass
at room temperature but will fail at elevated temperatures.
This same reason is why so many, instead, recommend more
fans. More fans to cure the symptom rather than use heat to
identify and replace a failing component.
In the meantime, what does the system (event) log report?
Does Device Manager report any conflicts or other problems?
See Windows Help, if necessary, to locate these two important
sources of information.
Measure DC voltages from power supply. A motherboard
voltage monitor is not sufficient for measuring voltages.
That voltage monitor can only report voltage changes - not
accurately report voltages. All voltages must measure within
upper 3/4 limits of voltage chart as posted in either:
"Computer doesnt start at all" in alt.comp.hardware on 10
Jan 2004 at
http://tinyurl.com/2t69q or
"I think my power supply is dead" in alt.comp.hardware on 5
Feb 2004 at
http://www.tinyurl.com/2musa
Essential is a 3.5 digit multimeter; so ubiquitous as to be
sold even in Sears, Home Depot, Lowes, and Radio Shack. That
is the least expensive solution to measuring voltages and a
tool as important as a screwdriver.
BTW don't waste good money on a UPS or surge protector to
fix this problem. First, surges occur typically once every
eight years. Second neither the plug-in UPS nor power strip
protector even claim to protect from surges that damage
hardware. Third, if AC mains voltage dropped so low as to
cause a computer failure, then incandescent lamps are
routinely dimming to less than 40% intensity. Furthermore, if
such AC mains voltage problems were causing problems, then the
multimeter would identify that problem. Brownouts never
damage properly constructed electronics. Brownouts are
destrutive to motors. But computers are more resilient.
Colin Cogle has recommended some good software solutions.
But his knowledge of the hardware side is severely lacking.
He recommends too many hardware myths. 1) Surges destroy
hardware. How many other appliances are damaged? 2) Spikes
are so irrelevant to computers that spikes are routine from a
UPS in battery backup mode (270 volt spike on the 120 volt
output). Spike may damage small electric motors but will not
harm the more resilient computer. Spikey outputs are one
reason why plug-in UPSes are so inexpensive. They can output
large spikes because computers are a most resilient
appliance. 3) Brownouts never damage properly constructed
electronic hardware. In fact Intel specs demand that a
computer even power up normally when AC voltage is so low that
incandescent lamps are at 40% intensity.
Spikes, surges, and brownouts causing problems are too often
promoted by myths. If any power line problem is causing your
problem, then the procedures cited above using a 3.5 digit
multimeter will identify that problem - in seconds.
I cannot emphasize enough the power of a tool called heat in
locating intermittent and future failures. There is no less
expensive tool than a multimeter to confirm electricity from a
power supply. Once you have used the multimeter to calibrate
a motherboard voltage monitor, then that voltage monitor can
properly report future voltage problems. But again, first the
multimeter is necessary.