NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/sci/electron/18350

Internet Message Format | 1992-11-04 | 3.8 KB

Xref: sparky sci.electronics:18350 misc.forsale.computers.pc-clone:599 misc.forsale.computers.d:273
Path: sparky!uunet!ogicse!news.u.washington.edu!stein.u.washington.edu!bketcham
From: [e-mail] (Benjamin Ketcham)
Newsgroups: sci.electronics,misc.forsale.computers.pc-clone,misc.forsale.computers.d
Subject: Re: 486 CPU KOOLERS FOR SALE
Message-ID: <[e-mail]>
Date: 5 Nov 92 08:31:10 GMT
Article-I.D.: u.1992Nov5.083110.7351
References: <[e-mail]> <[e-mail]> <[e-mail]>
Sender: [e-mail] (USENET News System)
Organization: University of Washington, Seattle
Lines: 56

Well, I can't say I've done any tests on CPU lifetime vs. heat, but that
doesn't seem to have stopped many of the others who have posted in
regards to this CPU fan thread, so why should it stop me?

Everyone seems to agree that lower temperatures are in general better
for semiconductors.  However, the question seems to be, does the decrease
in CPU temp that these fans achieve really make any real difference?  I
find it hard to believe that it does.  I have actually *never* seen a
CPU fail in my relatively small personal sample of systems...

CPUs just aren't the weak link in a computer system.  The
things that will fail will be the disk drives, connectors and sockets,
switching power supplies, monitors, and keyboards.  Statistically, as
memory gets denser and larger, the chances of radiation hits causing soft
errors will start to become a bigger problem for personal computers, as
it already is for larger systems.  But CPU failures?

My guess is that
90% of the small number of CPUs that ever do fail do so due to power
spikes, esd, transients from the bus or the failure of other chips, or
(probably the biggest factor), incorrect installation by unqualified
users.

Semiconductors are rated to operate normally at temperatures that
subjectively feel "too hot" to the touch (85 C) without any compromise
of reliability.  If these 486s are really exceeding 85 C, then maybe
these fans are needed, but I really doubt it.  (Yes, I have a source
for this 85 C number, albeit a questionable one I'll admit:  the 1986
Intel 80386 data sheet.  (That's case temperature, not ambient or
junction temperature.)  So maybe, just maybe, the 486 is rated for a
lower operating temperature, but I really doubt that, too.)

Plus, nobody has considered the possibility that these fans might do
more harm than good.  For one thing, as any service technician would
agree, I'm sure, the biggest problem impacting system reliability is
often the users opening the box up.  The chances of slipping with a
screwdriver and doing major damage are probably orders of magnitude
higher than the chances of a CPU failing within even 10 years.  What
about the vibrations and electrical transients introduced by these
fans?  Were any studies done about these factors?  After all, they are
bonded right on top of the CPUs.  I would imagine that the CPU sockets
are probably a good bit more susceptible to failure or intermittent
noise problems, exacerbated by vibration, than the CPU itself.

So for those few critical applications where even miniscule chances
of failure are a big problem, just replace the commercial-grade 486
with a mil-spec temperature range one, and forget about heat problems.
And the rest of us can forget about them too, I'll wager.  If someone
wants to do something that has a real impact on computer reliability,
develop mass storage devices without moving parts, and work on better
error detection and correction for RAMs.  The only good reason I have
heard of for cooling a CPU (with a Peltier device, in this case) is
to run it twice as fast as it could run at normal operating temperature.
Temperature-induced failures?  I don't think so.

--ben