[link|http://www.pcpowercooling.com/home.htm|PC PowerCooling]

These guys have been around since the 8086. Believe I first heard of them from Jerry Pournelle (!). Dunno how their QC has held up under competition from Bangladesh and the ascendance of the PHB in all of Murican 'what-passes-for' life, but they *were* first-rate design and materials.

[ps ignorance rant]
I cannot imagine.. putting a 'server' into operation for something more important than saving old Freecell scores -- without the Best PS available, and also *a bloody Working Temp Alarm* with multiple sensors for HD, CPU, enviro of case... All now so *cheap* to do - too. But: an extra step.

But then... I ain't no IT Pro: I actually understand something about electronics, heat, ripple - cheap aluminum electrolytics often substituted: NOT rated "low-ESR" (equivalent series resistance, that is). These especially for the lo-voltage sub-circuitry for those GHz goodies... thus on the MB and hard to envision wholesale replacing through those 4, 6? layers. Cheap mobo? for a Server ?!

Imagine what combos of the above factors can do to: er MILLIONS of logic-level switches, as that old 120 Hz or much likelier (switching freq) sawtooth + glitches - comes and goes. Odd reboots (even for Billyware?). "Can't write to disk 0" ?

It always amazes me that, people will entrust a (once) $1K CPU and a (once) $500 HD, $350 video card to: a $8 (qty) China PS. And never worry pretty little heads as in, "what's a power supply?"

Me? even for my el-cheapo Korean PS in the e-Machine: I at least measured the ripple; looked with a scope -- and will do so at least as frequently as thinking about CMOS battery. Scope optional - decent Fluke RMS meter will do = under load = FULL load. (make a CD from another via HD buffering and calculating Billy's fortune in background - OK compile a kernel too, in that Other OS) Then you *know* (at least today).

[/rant]

Yeah.. it's just One More thing to add to the list, "are my IIS patches checked.. since last week?"



Ashton Crank

In past 5-10 years, cost of making a decent root-mean-square responding meter has come way down. This is important when looking for the kinds of interference as would screw up digital logic (particularly at the lower and lower voltages). So you want to read that "spike" with reasonable accuracy - though you'd understand a lot more, if you saw its waveshape on a scope, too.

We lazy folk look at it this way: would you rather set up a battery of utilities; see if you can sleuth out mysterious digital events - or just see first, if there's a simpler explanation? (I'll bet lots of good mbs have been tossed.. and lots of wasted hours running logic tests, because of bad PSs.)

And no, Fluke isn't the only choice - but RMS is pretty important.. and a high-enough freq. response as matches the switching freq. - in order to "believe".


A.