Mayayana wrote:
"Paul" wrote

| All IDE/SATA drives now have SMART.
|
| It provides statistics.
|

I understand that much. When I researched it in the past
it never seemed to be very useful. The experts seem to
say that the numbers require careful interpretation, and
even then are misleading. So what's the point if it can't
be depended on?

You seem to have looked into the details far more than
I would ever think to. I don't think I've ever actually had
a drive die. And I've repaired many cheap PCs for friends,
with drives 10-12 years old and still going. Typically I'll
replace those and save the old for backup.

Now I have a 3-year-old WD Blue that the diagnostic
software says is kaput. I'm wondering if there's any
validity to Neil's belief that Win8 may have somehow killed
it, and if so, how that might be possible.
I also wonder if it may currently be typical to have
software installed that never stops accessing the disk.

In other words, my concern is not with the efficiency
of SMART in predicting a dying drive. I wouldn't want
to depend on that anyway. I'm just wondering whether
there may be special considerations with the "Metro
Series" of Windows.

Three years is the typical low-end prediction of
a drive's life, but I've never actually seen one die
so soon.

If you apply a pathological load to a regular
hard drive, it lasts around 1 year. This is a number
reported by people running web servers, where the disk
never stops moving the heads.

For example, if you wanted to try that at home, you might
store a million test files on the disk, such that they
span the entire surface, then use your random number
generator and ask the disk to read files at random.
That will throw the head around from inner to outer
ring. It flexes and un-flexes the actuator cable. And
wears the bearing the actuator rests on. That bearing
is not "frictionless" like the bearing inside the FDB
(fluid dynamic bearing) motor.

The flex cable is actually specifically designed for
each drive. On drives that short-stroke (such as perhaps
a 15K drive providing 300MB/sec transfer rates), the cable
will have different design requirements than your 7200RPM
drive where the arm moves the full distance over the
surface. You can't even interchange internal flex cables
between drives, because they're optimized for how the
drive works. Maybe the mass of the actuator arm, is
part of the equations.

Now, what is the drive in your laptop doing ? Well, it
doesn't have the pathological condition applied to it.
On drives that spin down, you might even notice on
occasion, that the heads park. And that's an indicator
that there isn't constant access. The drive is rated
for 300,000 head parks.

I have had a constant access situation. The optimizer
that moves prefetch files around or something, got
"stuck" one day, and using ProcMon, I could see constant
read and write to the same sector (basically rewriting
the sector with its current contents) over and over again,
at max speed. I could hear a high pitched "singing" sound,
even though the heads don't need to be moved around
when doing that. Just doing a little bit of regular
defrag, caused the prefetch optimizer to stop doing that :-)
Now, that's custom code in WinXP era, and entirely
different ("written from scratch") code is used in
Windows 8.

There is unlikely to be that kind of activity on
Windows 8, but between perfmon.msc and ProcMon.exe
(Sysinternals) you could probably figure it out.

And always remember, that the instrumentation on the OS
is incomplete. The OS is not fully transparent when it
comes to logging hardware activity. Many times, I'll
instrument a situation, and a graph will read zero,
when I know for a fact, there is activity. Like just
yesterday, I couldn't see some pagefile activity...

Paul