Ian & Stuart's Australian Mac 1993 September/September 93.iso/Archives/Mac FAQ/to off or not to off

Internet Message Format | 1993-10-11 | 12KB

Date: Sat, 11 Sep 1993 11:07:50 -0800
From: [e-mail] (Shannon Thornburg)
Subject: Macintosh Power Off/On Article

Several people have requested copies of the article by Tim Oey on turning
Macs on/off and how it affects computer lifetime.  This is information that
everyone who owns a Mac should look at, since it gives guidelines on how
you can make your computer last as long as possible.  Tim has generously
agreed to make the information available electronically.

Here is the latest version of that article, in text format.  See the notes
at the end for more information on distribution guidelines.


(Moderators:  this should probably be archived as something like
info-mac/info/hdwr/power-cycle.txt or off-not-off.txt)



ARTICLE BEGINS HERE

To Off or Not to Off, That Is the Question

by Timothy S. Oey

Over the years, there has been much controversy about whether it is better to
turn your computer off when it isn't used or leave it on continuously. This is
not a simple question to answer, and as it turns out there are many different,
but valid, answers. For the purposes of this discussion, we'll assume that
"better" means that the computer will last longer.

The simple answer to this question is: It's usually best to turn the computer
off whenever it won't be used for 8 hours or more. So if you use your computer
frequently during the day, as many do at work, turn it on in the morning and
off at night. If you use your computer less (a home computer for example), then
it is even more advantageous to turn your machine off. Besides, in both cases
you'll save energy.

Now if you'd like a more complete answer, or would like to know the details
behind the above conclusion, please read on.

The following information was distilled from numerous conversations with
engineers at Apple Computer, Conner Peripherals, and Quantum Corporation. The
conclusions reached are not necessarily those of any of these companies or the
engineers, but they are an attempt to derive information which the average
computer user should find helpful.

To maximize the total number of successful operational hours for any computer
device (i.e., maximize power-on hours), all computer devices (hard disks, CPUs,
monitors, other electronics) should be turned on and left on forever until they
fail. The number of power-on hours is what most engineers measure, and so
they'll tell you to leave your computer on all the time to maximize them.

But this does not necessarily mean that you, the user, will maximize the amount
of productive time you get out of the computer. For instance, during the night
the computer may be on but it may not be accomplishing anything for the user.
Let's call this productive time the user's perceived system life span. It's the
span of time over which the user is getting useful work out of the computer.
Users are probably more interested in maximizing their computer's perceived
system life span than the actual number of power-on hours.

Regardless of the wear caused by turning your computer on and off, there are
other factors that can have a much greater impact on your computer's life
expectancy. It is very important to treat your computer with care. Although not
the focus of this article, here is a brief list of common sense dos and don'ts
that will help ensure a lasting and worthwhile relationship between you and
your computer:
1. Keep it in a dust/dirt/smoke-free environment.
2. Don't spill things on it. Keep it dry.
3. Use a surge protector.
4. Don't drop it or jar it severely.
5. Use a screen saver to prevent monitor burn-in.
6. Keep it cool (room temperature) and out of the sun.
7. Don't block its ventilation slots.
8. Back up your files -- all systems are guaranteed to fail sooner or later.

Before we go on, be reminded that reliability is a complex probabilistic
science. Yes, probability and statistics can be tricky, but they are necessary
for determining when a computer is likely to fail. Manufacturers often use the
term MTBF (Mean Time Between Failures) as an indication of reliability. This
means that, given a large number of computers, the average one will run X hours
before a failure occurs -- X being its MTBF. This does not mean that yours or
mine will last X hours. It only means that ours will probably last about that
long. There is no way that a manufacturer can determine exactly how long any
given computer will last, just as it is not possible to determine exactly how
long you will live, assuming normal circumstances.

Manufacturing defects tend to be the single largest cause of computer hardware
failure. Manufacturers cannot guarantee that every device they make is perfect.
However, some are better at minimizing the number of problems that occur.
Apple(R) Computer, for instance, does an amazing amount of testing on all of
its computers and peripherals, as well as keeping close track of failure rates
in the field. Most well-known manufacturers do a good job in terms of design
and manufacturing. You will nearly always be better off, in terms of
reliability, if you buy equipment from reputable manufacturers. This cannot be
stressed enough.

The whole point of buying a computer is to accomplish useful work. Be careful
of the tradeoff between cost and the amount of useful work you will be able to
get from your computer. The same goes for disk size, CPU speed, number of
colors, etc. These don't necessarily mean that you will maximize the utility
you get from your computer. Yes, there are worthwhile bargains out there, but
you get what you pay for.

Many believe that turning a computer system on is the primary reason for
failure, because most systems fail at this time. This is not necessarily true.
It's just that this is the most likely time for weaknesses to become apparent,
because this is when the computer system undergoes its greatest stress.
Turning your computer on and off regularly (as prescribed above) may be even
more advantageous than it first appears, especially in the first year of
ownership. The stress of turning a machine on and off makes it more likely that
any manufacturing defects will become apparent sooner rather than later -- and
hopefully before the warranty runs out. Once manufacturing defects are weeded
out, it is highly likely that a computer will run for quite a while before wear
causes a failure.

Assuming the average benign environment, the most significant causes of wear,
in rough order from most to least, are:
1. Heat
2. Power cycling (turning a machine on/off)
3. Power-on hours
4. Humidity/salt/airborne pollutants
5. Age (yes, some components incur wear even when not used)

Heat is a problem because electronic components may burn out if not cooled
sufficiently. Excessive heat can damage any component, especially physically
moving ones such as disk drives. Power supplies are sometimes a bit erratic for
the first few microseconds when first turned on, resulting in initial power
surges. Heating/cooling cycles can cause joint failures due to differing
expansion properties between materials. The various wear factors have the
greatest impact during power-on hours. Monitor phosphors and filaments
eventually burn out. Humidity, salt, and pollutants can corrode various parts.
Simple aging can also have an effect, although this is very small compared with
the others. There are countless other effects related to the above causes -- to
many to enumerate here.

Note that turning computer systems on and off vs. leaving them on may not
really matter much for today's average user buying new equipment. Computer
equipment is increasingly well engineered and reliable. It is much more likely
to become obsolete than wear out.

Hard disk drives, for instance, are by far the most likely component of a
computer system to wear out because they are mechanical and undergo fairly high
stress. Most of the other components -- power supplies, monitors, logic boards,
other electronics -- last significantly longer in comparison. Since hard disks
are the weakest link in the computer, we'll focus on them for the remainder of
our discussion.

Most high-quality hard disk drives are rated for an average of 20,000 on/off
cycles and an MTBF of 25,000 hours or more. If you turn your machine on/off
once a day, it will take 55 years to reach 20,000 cycles. An MTBF of 25,000
hours means that the average hard disk should last about that long, and 25,000
hours is equal to 2.9 years of non-stop running. Again, remember that these
numbers reflect probabilities, not certainties. In testing, hard disks
sometimes survive 100,000 on/off cycles and the equivalent of 1,000,000 hours
of continuous operation. On the other hand, some fail much sooner.

<<1993 Update: Since the time this article was first written (Feb 1990), hard
drive reliability has increased.  MTBF ratings for most drives (as of 1993) are
now in the 250,000 to 350,000 hour range. However, the basis for calculating
these hours appears to have changed from an MTBF based on power-on hours to an
MTBF based on total hours (on and off). The drive manufacturers may have
incorporated typical use patterns into their calculations to boost the MTBF
number and perhaps give people a number that on average is more meaningful --
250,000 hours translates to 29 years of average use.  Regardless, while hard
disk drives may or may not be the weakest link in the computer anymore, and
while MTBF ratings for other components may or may not have increased
correspondingly, the logic above and below remains valid even though some
numbers may have changed.>>

Now the question is: How much wear does turning a system on and off really
cause? This is something that no one seems to have calculated yet. In fact,
most will not even hazard a guess. However, let's take a crude and somewhat
pessimistic guess that relates on/off wear to power-on-hours wear for
comparative purposes. Let's say that the wear caused by turning a machine on
and off is roughly equivalent to that caused by 8 power-on hours.

To do a rough calculation of how long a system will last if we turn it on and
off each day, let's assume that a business computer is on for 8 hours each day,
5 days a week, 52 weeks a year, and that power-off time causes essentially no
wear. Let's also assume that the MTBF for our computer is about 25,000 hours
(this may be optimistic for a whole system, but it's good enough for our
purposes). This yields the following equations, where <life span in weeks> is
the user's perceived life span for the computer system:

<on/off wear in hours>
= <life span in weeks> * 5 cycles/week * 8 hours/cycle
= <life span in weeks> * 40 hours/week

<power-on wear in hours>
= <life span in weeks> * 5 cycles/week * 8 hours/cycle
= <life span in weeks> * 40 hours/week

<<on/off wear in hours> + <power-on wear in hours> = 25,000 hours

<life span in weeks> * 40 hours/week + <life span in weeks> * 40 hours/week =
25,000 hours
<life span in weeks> * 80 hours/week = 25,000 hours
<life span in weeks> = (25,000/80) weeks = 312.5 weeks = 6.0 years

If we leave the computer on continuously:

<on/off wear in hours>
= 0

<power-on wear in hours>
= <life span in weeks> * 24 hours/day * 7 days/week
= <life span in weeks> * 168 hours/week

<on/off wear in hours> + <power-on wear in hours> = 25,000 hours

0 + <life span in weeks> * 168 hours/week = 25,000 hours
<life span in weeks> = (25,000/168) weeks = 148.8 weeks = 2.9 years

Notice that the first case yields twice the life span of the second case,
although the actual number of successful power-on hours is halved. A computer
used less frequently would yield even better results.

Bottom line: Turn your system off when you won't be using it for 8 hours or
more. But for the most part don't worry about it, because if you bought your
computer system from a reliable manufacturer, it will probably last a very long
time without any hardware failures. It is much more likely to become obsolete
than it is to wear out. Remember that all systems will fail eventually, so keep
backups.


Copyright (c) 1990, 1993  Timothy S. Oey.  Tim Oey is a project manager at
Apple Computer and the Forum Leader for BikeNet on America Online.  He can be
reached at [e-mail], [e-mail] or [e-mail].  Permission is
granted to distribute this article electronically for free as long as it
remains a complete whole.  Please contact the author if you wish to re-publish
the article in some other form.