Power On Time Count SMART warning for Fujitsu 20GB

[Franc Zabkar]

My MPF3204AT Fujitsu 20GB HDD's Power On Time Count just triggered a
SMART warning after dropping below the threshold of 20.

The raw value is currently 43763894 seconds, ie 12156 hours.

OTOH my 120GB Seagate drive is currently at ...

Power On Hours Count 9 0 99 99 0000000005DDh (1501)

.... so its expected life appears to be somewhere between 150,100 and
75,050 hours.

Are these numbers realistic, ie has hard disc reliability really
improved that much? I don't think so.

The Fujitsu datasheet ...

http://www2.fcpa.fujitsu.com/sp_support/ext/desktop/datasheets/mpf3xxxat-datasheet.pdf

.... specifies the number of start/stop cycles as 40,000.

A log of SMART reports shows the following:

Start/Stop Count 4 16 98 98 000000000578h EC
Start/Stop Count 4 16 97 97 00000000057Dh EC

This suggests that each point represents about 468 starts, and a loss
of 84 points represents 39,340 start/stop cycles which roughly tallies
with the spec.

- Franc Zabkar

 

[Arno Wagner]

My MPF3204AT Fujitsu 20GB HDD's Power On Time Count just triggered a
SMART warning after dropping below the threshold of 20.

The raw value is currently 43763894 seconds, ie 12156 hours.

OTOH my 120GB Seagate drive is currently at ...

Power On Hours Count 9 0 99 99 0000000005DDh (1501)

... so its expected life appears to be somewhere between 150,100 and
75,050 hours.

Are these numbers realistic, ie has hard disc reliability really
improved that much? I don't think so.

This is not reliability. This is component life under operation.
It has improved significantly due to better lubrication
material as has non-operating component life. Here also because
of the move to ceramic capacitors, that basically live forever,
while electrolytes die after 3-5 years or so, depending on
temperature and quality.

But, no, I would say the 7.5/15 years are probably still
longer than you can realistically expect. Typical component
life for industrial electronics is still around 5 years.
I guess this is just so that the range is large enough.
It should be an "old age" attribute anyways, not a "pre-fail."

The Fujitsu datasheet ...

... specifies the number of start/stop cycles as 40,000.

A log of SMART reports shows the following:

Start/Stop Count 4 16 98 98 000000000578h EC
Start/Stop Count 4 16 97 97 00000000057Dh EC

This suggests that each point represents about 468 starts, and a loss
of 84 points represents 39,340 start/stop cycles which roughly tallies
with the spec.

Well, you should be prepared for your drive to become unreliable
in the not to far future. But these values are not hard. If you
have backup, you can continue to use them and may even get some
more years of reliable operation. And, realistically, that is not
too different from what a new deive gives you.

Arno

 

[Rod Speed]

My MPF3204AT Fujitsu 20GB HDD's Power On Time Count just
triggered a SMART warning after dropping below the threshold of 20.

The raw value is currently 43763894 seconds, ie 12156 hours.

OTOH my 120GB Seagate drive is currently at ...

Power On Hours Count 9 0 99 99 0000000005DDh (1501)

... so its expected life appears to be somewhere between 150,100 and 75,050 hours.

Are these numbers realistic,

Nope, but not for the reason you are going on about.

ie has hard disc reliability really improved that much? I don't think so.

Fraid so.

The Fujitsu datasheet ...

... specifies the number of start/stop cycles as 40,000.

A log of SMART reports shows the following:

Start/Stop Count 4 16 98 98 000000000578h EC
Start/Stop Count 4 16 97 97 00000000057Dh EC

This suggests that each point represents about 468 starts, and a loss of 84
points represents 39,340 start/stop cycles which roughly tallies with the spec.

But that is a design value. There is no suggestion that the drive does deteriorate uniformly as the start/stop count
increases.

 

[Simon]

This is not reliability. This is component life under operation.
It has improved significantly due to better lubrication
material as has non-operating component life. Here also because
of the move to ceramic capacitors, that basically live forever,

while electrolytes die after 3-5 years or so, depending on temperature and quality.
Fantasy.

But, no, I would say the 7.5/15 years are probably still
longer than you can realistically expect. Typical component
life for industrial electronics is still around 5 years.

Like hell it is.

I guess this is just so that the range is large enough.
It should be an "old age" attribute anyways, not a "pre-fail."

Well, you should be prepared for your drive to become unreliable in the not to far future.

Mindlessly silly.

But these values are not hard. If you have backup, you can continue
to use them and may even get some more years of reliable operation.

Gets sillier by the minute.

And, realistically, that is not too different from what a new deive gives you.

Even sillier.

 

[Arno Wagner]

Fantasy.

Read a datasheet.

Like hell it is.

Again, read a datasheet.

Mindlessly silly.

And again, read a datasheet.

Gets sillier by the minute.

And get a clue about device live statistics.

Even sillier.

You have no clue. On a guess I would say one of the usual clueless
bigmouths, hiding behind a new name, because nobody
listens anymore.

Arno

 

[Simon]

Read a datasheet.

Doesnt say a damned thing about that mindlessly silly claim.

And have fun explaining how come the power supplys which have many more
much bigger electros in them dont even have that 3-5 years in their datasheet.

Again, read a datasheet.

Again, have fun explaining the power supply data sheets say nothing like that
and that fact that decent power supplys last a lot longer than that anyway.

And again, read a datasheet.

And again, the datasheet says nothing like that on the drive becoming unreliable.

And get a clue about device live statistics.

It you that needs to do that.

You have no clue.

You in spades.

On a guess I would say one of the usual clueless bigmouths,
hiding behind a new name, because nobody listens anymore.

Guess again. Those who reply must have done, fool.

 

[Squeeze]

Franc Zabkar wrote in news:[email protected]

My MPF3204AT Fujitsu 20GB HDD's Power On Time Count just triggered a
SMART warning after dropping below the threshold of 20.

The raw value is currently 43763894 seconds, ie 12156 hours.

OTOH my 120GB Seagate drive is currently at ...

Power On Hours Count 9 0 99 99 0000000005DDh (1501)

... so its expected life appears to be somewhere between 150,100 and
75,050 hours.

Based upon what?

 

[Squeeze]

Arno Wagner wrote in news:[email protected]

Read a datasheet.

Maybe there are no electrolytic capacitors in my 25 year old TV then.
Or perhaps they are of milititary spec, right, Babblebot?

[snip]

You have no clue.

But you have, Babblebot. Other may call it a braintumor though.

On a guess I would say one of the usual clueless bigmouths,

hiding behind a new name,

Unlike you, eh Babblebot.
Clueless, big mouth, but hey, at least you use your own name.

because nobody listens anymore.

Would that nobody be you, babblebot?
Courtesy of your extensive kill file?
Making you conveniently deaf.

 

[Arno Wagner]

Arno Wagner wrote:

Umm... Arno, many (most?) of us have first-hand experience with
electronics lasting MUCH longer than that. And of they all, in
general, use electrolytic caps.

In my experience, 10+ years of lifetime is routine and, really,
expected from an electronic device.

Remember the "capacitor plague"? There the lifetime was reduced
from about 5 years average to 1-3 years average on typical operating
conditions. Where you hit? I was several times, the first one after
2 years operating time of the mainboard in question.

Then there is the thing, that the life time is not the time
after which it fails, just the time where the probability per
time unit starts to increase significantly. Of course some units
will life much longer. I have several HDDs older than 5 years
that work fine. Also the failure mode can vary greatly.
For example, if you have a good quality PSU, probably even a total
capacity loss of an input filter electrolythe will not matter much.
With a bad PSU it may lead to significant problems.

Also, we are talking about a HDD here. It may run hot. Derate
lifetime typically at a factor of 2 per 10C. A well cooled PCB
may be at 25C. An uncooled at 55C or more. I did not only say
"3-5 years" , I added "depending on temperature and
capacitor quality". I should also have added "design quality".

What I claimed for capacitors was that circuit designers
tend to target a 5 year lifetime at _expected_ contitions.
Most people here will cool HDDs well. That are not the
expected conditions anymore. But remember all the people
that lost Maxtors early and it turned out they were not
cooled well? Sometimes after 1-2 years?

So predictiong lifetimes is difficult. Possibly I should
have been far more specific. I apologize for that. But
the numbers are not fantasy at all.

Here is a real number from a very high quality low-ESR
Electrolythe capacitor (Rubycon ZLH, 7mm hight): 1000h at 105C.
Lifetime here means capacity within +/-25% of original, leakage
still within datasheed value and dissipation (related to internal
resistance) <= 200% of datasheet value.

Arno

 

[Arno Wagner]

chrisv wrote:

If you want, you can come-over and I'll fire-up my circa 1985 Amiga
A1000 for you. (But I conceed that the HD is newer - circa 1989 8).

No need, I have an Atari ST that works fine. But this
is not "lifetime". This is "shelf live unoperational". And that
Atari ST has some replaced electrolythes, because I had to fix them
to solve a stability problem.

Alas, my (also) circa 1985 Sony KV25XBR died within the last year,
after over two decades of heavy use...

Good design then. My (expensive) Sony Vaio died after 2 years of
light use. The cause was chipset overheating due to inadequate
cooling.

It just boild down to several things:

- Treat electronics well and they will live longer
- It is statistics. Lifetime is just when failures start to get more
likely. Some things die withing a year, some keep 20 years. Can
still be 5 years lifetime.
- Not all failures kill a thing. Some do not matter at all.
- Personal experience is not a global predictor.

Arno

 

[Simon]

No need, I have an Atari ST that works fine. But
this is not "lifetime". This is "shelf live unoperational".

Plenty have had TVs work fine for that long, including me.
And that isnt shelf life unoperational, thats daily use.

And that Atari ST has some replaced electrolythes,
because I had to fix them to solve a stability problem.

And most of us have had TVs that havent needed any electos replaced.

Good design then. My (expensive) Sony Vaio died after 2 years of
light use. The cause was chipset overheating due to inadequate cooling.

It just boild down to several things:

- Treat electronics well and they will live longer

Treat a properly designed TV normally and it will last a hell of lot longer than your mindlessly silly claim at the top.

- It is statistics. Lifetime is just when failures start to get more likely.
Some things die withing a year, some keep 20 years. Can still be 5 years lifetime.

Not when the vast bulk of TVs do a lot better than 5 years.

- Not all failures kill a thing. Some do not matter at all.
- Personal experience is not a global predictor.

Decent stats are tho.

 

[Simon]

Arno Wagner wrote

Umm... Arno, many (most?) of us have first-hand experience with
electronics lasting MUCH longer than that. And of they all, in
general, use electrolytic caps.

In my experience, 10+ years of lifetime is routine
and, really, expected from an electronic device.

Yeah, he's never had a clue about the basics.

 

[Simon]

Remember the "capacitor plague"?

Its irrelevant to that claim of yours at the top.

There the lifetime was reduced from about 5 years average

You've plucked that from your arse. We can tell from the smell.

to 1-3 years average on typical operating conditions. Where you hit?

I have had just one device affected out of dozens that werent.

I was several times, the first one after 2 years operating time of the mainboard in question.

Irrelevant to that stupid claim of yours at the top.

Then there is the thing, that the life time is not the time
after which it fails, just the time where the probability
per time unit starts to increase significantly.

Not generally a problem with modern hard drives.

Of course some units will life much longer. I have
several HDDs older than 5 years that work fine.

In fact of all of mine, only 1 doesnt.

Also the failure mode can vary greatly. For example, if you
have a good quality PSU, probably even a total capacity
loss of an input filter electrolythe will not matter much.

Wont matter at all in fact. So that claim of yours at the top is mindlessly silly.

With a bad PSU it may lead to significant problems.

Also, we are talking about a HDD here. It may run hot.
Derate lifetime typically at a factor of 2 per 10C.

Another number straight from your arse. We can tell from the smell.

A well cooled PCB may be at 25C. An uncooled at 55C or more.
I did not only say "3-5 years" , I added "depending on temperature
and capacitor quality". I should also have added "design quality".

You also made stupid claims about the datasheet.

What I claimed for capacitors was that circuit designers
tend to target a 5 year lifetime at _expected_ contitions.

No they dont, most obviously with power supplys.

Most people here will cool HDDs well. That are not
the expected conditions anymore. But remember all
the people that lost Maxtors early and it turned out they
were not cooled well? Sometimes after 1-2 years?

Irrelevant to hard drives in general.

So predictiong lifetimes is difficult.

And your claim right at the top is mindlessly silly.

Possibly I should have been far more specific. I apologize
for that. But the numbers are not fantasy at all.

The numbers at the top are pure fantasy.

Here is a real number from a very high quality low-ESR
Electrolythe capacitor (Rubycon ZLH, 7mm hight): 1000h at 105C.
Lifetime here means capacity within +/-25% of original, leakage
still within datasheed value and dissipation (related to internal
resistance) <= 200% of datasheet value.

Irrelevant to hard drives where that doesnt matter.

 

[Arno Wagner]

Plenty have had TVs work fine for that long, including me.
And that isnt shelf life unoperational, thats daily use.

And that is a TV, not a HDD and has different engineering
parameters.

And most of us have had TVs that havent needed any electos replaced.

See above.

Treat a properly designed TV normally and it will last a hell of lot longer than your mindlessly silly claim at the top.

No argument from me. But the TV set comparison is completely
besides the point.

Not when the vast bulk of TVs do a lot better than 5 years.

Huh? Did you read the sentence I wrote?

Decent stats are tho.

Indeed.

Arno

 

[Arno Wagner]

Its irrelevant to that claim of yours at the top.

You've plucked that from your arse. We can tell from the smell.

You did obviously not read an analysis of that problem and the
normal situation it is compared with.

Consult the IEEE Spectrum, for example.

Incidentially *plonk* for incompetence, baseless insults
and a big mouth.

Arno

 

[Simon]

And that is a TV, not a HDD and has different engineering parameters.

He said AN ELECTRONIC DEVICE at the top and
clearly more than just hard drives are being discussed.

And you yourself made silly claims about electros, not just hard drives too.

See above.

See above.

No argument from me. But the TV set comparison is completely besides the point.

Nope, not when you made a completely silly claim about electros and even you should
have noticed that TVs and power supplys have a lot more of those than do hard drives.

Huh? Did you read the sentence I wrote?

Yep, the lifetime is clearly a lot more than 5 years when the vast bulk of them do a hell of a lot better than 5 years
and when they dont last that long, its usually the components which arent seen in hard drives that fail in that time.

Indeed.

And even you should have noticed that most hard drives last a lot longer than the 5 year design life SOME datasheets
specify.

 

[fang]

You did obviously not read an analysis of that problem
and the normal situation it is compared with.

Wrong, as always.

Consult the IEEE Spectrum, for example.

Have a look at what google reported on their own hard drives.

And then use a VERY large towel on your face.

Incidentially *plonk* for incompetence, baseless insults and a big mouth.

Fat lot of good that will do you, you stupid plonker.

 

[fang]

And that is a TV, not a HDD and has different engineering parameters.

He said AN ELECTRONIC DEVICE at the top and
clearly more than just hard drives are being discussed.

And you yourself made silly claims about electros, not just hard drives too.

See above.

See above.

No argument from me. But the TV set comparison is completely besides the point.

Nope, not when you made a completely silly claim about electros and even you should
have noticed that TVs and power supplys have a lot more of those than do hard drives.

Huh? Did you read the sentence I wrote?

Yep, the lifetime is clearly a lot more than 5 years when the vast bulk of them do a hell of a lot better than 5 years
and when they dont last that long, its usually the components which arent seen in hard drives that fail in that time.

Indeed.

And even you should have noticed that most hard drives last a lot longer than the 5 year design life SOME datasheets
specify.

 

[Franc Zabkar]

This is not reliability. This is component life under operation.

I don't see the distinction. The end user just needs to know whether
one device can be expected to last longer than another. If it does,
then it's more reliable. He doesn't care whether the difference is due
to design philosophy, or component selection, or the state of
technology.

It has improved significantly due to better lubrication
material as has non-operating component life. Here also because
of the move to ceramic capacitors, that basically live forever,
while electrolytes die after 3-5 years or so, depending on
temperature and quality.

But, no, I would say the 7.5/15 years are probably still
longer than you can realistically expect.

I tried to compare a ST313021A 13GB Seagate HD of around the same
vintage. Unfortunately Seagate's SMART data are confusing as usual.

Here is a sample:

Att Thr Cur Worst Raw
-------------------------------------------------------
Power On Hours Count 9 0 247 1 0000000026C7h
Power On Hours Count 9 0 210 1 000000002810h
Power On Hours Count 9 0 188 1 0000000028D0h

It looks like each point represents about 8.8 hours. A total of 256
points would then equate to about 2260 hours. However the raw figure
represents 10,448 hours.

As for the start/stop count, it appears that the threshold for a SMART
warning must be between 88,020 and 72,496.

Start/Stop Count 4 20 96 96 000000001131h
Start/Stop Count 4 20 96 96 0000000011B3h

However, the datasheet ...

http://www.seagate.com/support/disc/manuals/ata/u8pmb.pdf

.... specifies the number of "contact start-stop cycles" as 50,000.

- Franc Zabkar

 

[Arno Wagner]

On 4 Sep 2008 21:44:49 GMT, Arno Wagner <[email protected]> put finger to
keyboard and composed:

I don't see the distinction. The end user just needs to know whether
one device can be expected to last longer than another. If it does,
then it's more reliable. He doesn't care whether the difference is due
to design philosophy, or component selection, or the state of
technology.

True, but the problem here is if these things are not well-defined,
then vendors will lie and sell worse quality with seemingly better
numbers.

So the "component live" is actually the time period where the
steted MTBF holds (except initially). After the component live
time iis over, the MTBF gets worse, usually slowly.

I tried to compare a ST313021A 13GB Seagate HD of around the same
vintage. Unfortunately Seagate's SMART data are confusing as usual.

Here is a sample:

Att Thr Cur Worst Raw

It looks like each point represents about 8.8 hours. A total of 256
points would then equate to about 2260 hours. However the raw figure
represents 10,448 hours.

As for the start/stop count, it appears that the threshold for a SMART
warning must be between 88,020 and 72,496.

Start/Stop Count 4 20 96 96 000000001131h
Start/Stop Count 4 20 96 96 0000000011B3h

However, the datasheet ...

... specifies the number of "contact start-stop cycles" as 50,000.

That is pretty standard for non-notebook dsrives.

Arno