On Nov 17, 2009, at 10:27 AM, Bill Moran wrote:
[ ... ]
Not all power supplies are created equal. Unfortunately, there's
no government oversight on power supply ratings, thus a cheap 450W
power supply might go unstable if it has to supply 200W for very
long, whereas a good quality 200W power supply might be able to
put out 450W for short periods reliably.
A very good-quality power supply with a thermally activated circuit
breaker might tolerate a 250% overload for 20 seconds to a minute, but
anything with a fuse is likely to blow in some tens of
There are some widely used standards for computer power supplies;
almost all modern machines want ATX12V which is used by Intel P4s,
Core, etc and AMD Athlon, Athon64 platforms. Multicore boxes commonly
want another extension to the base ATX standard called EPS12V; both
are well-documented here:
The other major standard in 80-plus certification, which is linked to
Energy Star ratings; if you discount the branding, they still perform
functional tests of PSUs at 20%, 50% and 100% of rated load, and
confirm that the PSU isn't wasting excessive amounts of power. Saving
20-30 watts over time justifies the cost of a more expensive PSU, and
it doesn't hurt that the machine doesn't have to deal with the extra
thermal load. For example:
Any new PSU which isn't 80-plus certified is pretty likely to be
unable to run at 100% of rated load without failing.
Additionally, are you sure your service power is good? Even the
best power supply will fail if you're not getting 120V/60H at the
outlet (or whatever voltage/freq you're supposed to get in your part
of the world).
This is also a good point. If you know what you're doing and have a
multimeter, you can check your AC line and look for various issues
like voltage sag under load, current leakage to ground, etc. Failing
that, something like this Kill-A-Watt meter is quite handy:
...although, obviously, one would want to obtain a unit intended for
the local region's electrical standards if you are not in NA.
firstname.lastname@example.org mailing list
To unsubscribe, send any mail to "freebsd-questions-unsubscr...@freebsd.org"