[ltp] Re: T420s HD life and Load Cycle counter

[Henrique de Moraes Holschuh]

On Sun, 03 Jul 2011, Stefan Monnier wrote:
> >> Laoptop (and low power desktop drives) are designed to do such "load
> >> cycle" *very* often.  This number is unrelated to the number of
> > 100k to 300k cycles over the entire drive lifetime on the lesser devices,
> > 600k on the good stuff like Hitachi HDDs.  You can kill a heavy drive in a
> 
> Last I checked, the limit on "load cycle count" is never mentioned in
> specs, whereas the limit on spin-up/spin-down count (aka
> Start_Stop_Count) is and is indeed around 600k for many laptop drives.

Hitachi is explicit about it on all their datasheets.  And I just checked
the product manual for a current Seagate Momentus drive (ST9750423AS), and
it is in there too:  600k controlled unloads, 50k uncontrolled unloads
(happens when you cut power without getting the drive in a safe state
first).

> But admittedly, the specs are rarely very clear about this (e.g. the
> spec of my old desktop Seagate drive says "50K start stop cycles",

Laptop drives are engineered to survive a LOT more load/unload cycles (and
also spinup/down cycles) than desktop and server HDDs.  OTOH, they are most
definately not engineered for continuous operation like server HDDs.

It is quite probable that a desktop hdd of a given vendor would survive 50k
unloads while a laptop drive of the same vendor would survive 600k unloads.

> >> spin-up/spin-down; and other than FUD, there's no evidence that high
> >> numbers are a sign of problems to come.  Many manufacturers have simply
> > Head unloading slowly degrades the head support integrity, due to flex.
> > You can actually hear when the drive does it, btw: a very soft "click".
> 
> You can hear it on some drives and not on others.  The fact that you can
> hear it doesn't imply any kind of "evidence that high numbers are a sign
> of problems to come".

It is evidence of mechanical stress.

> > Which ones, so that I can avoid them like the plague?
> > Unloading heads is important for two reasons: unloaded heads are far less
> > succeptible to damage due to sudden impacts to the drive, and allow for the
> > linear head assembly positioning motor to be either turned off, or operated
> > in reduced current.
> > So, it reduces idle power consumption of the drive, which is of paramount
> > importance for the vendor's marketing department since everyone overstates
> > their battery longevity numbers.  It can also decrease the chance of a
> > damaged drive that needs to be returned in warrany on the cheap-o crap 1yr
> > warranty service.
> > Removing Load Cycle Count is just unethical.
> 
> Why?  Why would it be unethical to stop telling users how many times the
> head was loaded/unloaded?

Because you will not be made aware of activity that is detrimental to a
longer service life of the HDD *which you might not require in the first
place*.  Some of these devices are being programmed by the BIOS to unload
heads when idle for a few seconds.

I am not sure how good a parameter it is to track failure probability of the
device in the near or medium term, though.  The laptop HDDs that died on me,
did it well before they reached anywhere close to the load/unload cycles
they were rated for.  Most died due to spin motor issues (I don't shake them
much, so media problems and head damage is less likely in my usage pattern
:p).

It was not added in a whim in the first place, it should be removed only if
the underlying technology changed enough that it is not relevant anymore.
Keeping the customer in the dark is not an acceptable reason for removal.

AFAIK, spinning rust drives are not rated for >1M unloads just yet...

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh