[conspire] Preparing dual-boot system

Rick Moen rick at linuxmafia.com
Tue Jan 22 11:17:25 PST 2008

Quoting Ross Bernheim (rossbernheim at speakeasy.net):

> Over rating the supply by either outright untruths or tests that favor
> the strengths and ignore weaknesses of the power supply are quite
> common among the lesser known brands. Use cheap capacitors that will
> fail much sooner under stress. Use smaller capacitors so there is no
> margin of  safety.  Use smaller inductors and the supply won't work as
> well in a brown out or with surges or spikes.
> Cut the capacity of one or more of the many voltages a modern power
> supply provides. Reduce the number of regulators and tie some of the
> voltages to the main voltage. This is a subtle error as some of these
> voltages may sag when under temporary load while all else is fine.

Points well taken.  And remember, a PSU's task is two-fold:  (1)
Generate and furnish specified DC voltages to a variety of impedance
types (reactive loads being tougher on cheap PSU gear than are resistive
ones) up to the PSU's rated current limits.  (2) Protect attached
equipment from sundry glitches in your building's AC supply.

Whenever a PC's PSU falls short on either of those tasks, either
by overstressing from failure to handle attanched load competently,
or by passing along PG&E spikes, surges, over/undervoltage conditions,
and purely local AC wobbles (e.g., when your refigerator kicks on), 
it's highly likely to freak out and send destructive garbage-power to
that PC.

I figure about half the mysterious, unexplained failures of people's
motherboards, RAM, hard drives, etc. I hear about so frequently owe to
cheap, misbehaving PSUs, and (about) the other half to damage from
runaway heat buildup.

So, me, I prevent the first scenario by favouring PSUs that don't suck.
Some other people hide their cheapo PSUs behind uninterruptable power
supplies (UPSes), relying on the latter to protect the former, and I
notice that still often fails.  I use the same AC feed, and have no
such problems.  Their parts die young; mine don't.

The second scenario I avert by keeping airflow paths clear, clearing
away dust bunnies from vents, replacing fans (especially cheap ones with
sleeve bearings) before they seize up, eschewing nutso gamer gear, and 
favouring conservatively designed system cases that aren't running the
ragged edge of failure in the first place.

Both scenarios are so easily avoided, it's a real shame that so many
people get that part wrong.

> The biggest power guzzlers these days appear to be the top end Gamer
> graphics cards with multiple shaders/renderers and GPUs and the Gamers
> are now  tying multiple cards together! These are the guys who need
> 1KW or bigger  power supplies.  I won't say these people are crazy,
> just obsessed. But their needs  tend to warp the industry. Having the
> task of buying some computers for work, I found  it very difficult to
> find a PC that didn't have gaming graphic cards. Your only choice
> among the major suppliers is do you want a low end or high end gamer
> graphics card.

Right.  And that really _is_ crazy.  

You'd never know it from reading the Pee-Cee mags or looking at the
major VARs' systems, but I notice that Matrox (of Quebec) is still very
much in business -- and their video cards _always_ had much higher video
quality than the gamer-oriented crud that Nvidia and ATI produce.  What
a pity that the industry seems to have mostly forgotten them.  And,
since, increasingly, people's console systems are laptop (er, "notebook") 
computers rather than workstations, more often than not you're stuck
with what's integrated into the motherboard.  I guess, if you can't get
Matrox (and you're not a 3D gamer), you're probably better off favouring
integrated Intel video.

More information about the conspire mailing list