Wise to use HDD as offline storage?

[J. Clarke]

You'll be looking back when it turns out its just another
technology that never did manage to get up enough of
a head of steam to be viable over the longer term.

Uh, MOD has been around longer than DVD or recordable CD. And those first
disks are still readable in current generation drives.

Most say the same thing about MOD and when DVD continues to be
available and MOD doesnt, its you that will have wasted your money.

And when will that be?

Many feel the same way about technology that never does
get up enough of a head of steam to have a decent future.

MOD has plenty of "steam" in the professional market. Your argument would
apply equally well to, for example, LTO.

We'll see, when its no longer available by mail order either.

And when will that be?

Oh bullshit. You cant ignore DVD so glibly.

Actually, he can. MO does have a track record. Recordable DVD's is much
shorter and Arno does have first hand experience with DVD failures so he is
speaking from experience.

Its actually sinking beneath the waves, just like every other
technology that never did get up a decent head of steam.

It is? You have sales figures to support that argument?

And some avoid that fine using hard drives and DVD.

Depends on their needs. It's overkill for backup IMO but it's not overkill
for archival storage. The two are not the same.

 

[J. Clarke]

Arno - I'd like to understand what effect this point has. Is it intended
to mean that a given backup set may be unreadable at some future time
because you may be trying to use a DVD reader that is incompatible with
the DVD disc?

There are three issues of which I am aware. Apparently some writers have
difficulty with some media. If the writer vendor changes the internals in
midstream he might change to some that don't work well with the media
you're using. Different media chemistries have different degrees of
stability--if the media vendor changes chemistry in midstream without
telling anybody then the new media may be significantly less reliable than
the old media. DVD and CD drives can suffer alignment problems--a disk
that is quite readable in one drive may not be readable at all in another
that is apparently the same, even though both drives can read disks that
they wrote.

This leaves totally aside the format wars.

 

[Rod Speed]

1) XP original (SP0) can't "see" HDs < 137G

You've got that backwards. And they can see drives > 137G,
they'll just corrupt the data as access wraps at that point.

2) XP installations do not survive file-level copying, must image

Wrong. Have fun explaining xxclone.

And you dont need a bootable copy for backup anyway.

3) ATX "off" maintains power trickle; unsafe to remove IDE HDs

Wrong. And the worst you have to do is unplug
the power supply from the mains anyway.

4) S-ATA hot-swappability requires S-ATA power connection
5) Single recent backup may contain pre-existing problems
6) HDs don't sem to like brackets and exclosures
Bullshit.

So yes; you could use (say) BING from www.bootitng.com to image
off C: to another HD, as long as you unplug the PC from mains before
connecting and disconnecting the extra HD. Bear in mind the risk of
including problems within the backup, e.g. (say) a malware that
infected the system 29 days ago that has a 30-day payload fuse.

For the rest of the data, I'd do file-level copies rather
than image backups, so that the backups are browsable.

Any decent imager allows you to browse the image.

For the same reason, I'd keep data you wanted
to browse, off the C: that you will image.

No need. Just use a decent imager.

If copying files to CDR or DVDR, bear in mind that different file
system rules can knock file and path names around, attributes could
revert to read-only, and NTFS-specific metadata will typically be lost.

Another good reason for imaging with a decent imager.

To preserve those things, consider first archiving within a .zip
(using something that manages the NTFS fluff, if that's important
to you) and then drop those .zip onto the CDR or DVDR.

Or use a decent imager.

 

[Rod Speed]

So, means you should remove the power cord before removing hard drives?

He's just plain wrong and yes, you can remove the power cord.

Right. This 'simply' implies that you need to do a complete system
backup just after installing Windows and setting up your default
applications. Then, keep this known good backup separate from the
on-going/periodic backups, which can be used if you need to restore
from hardware failure.

Its better to have more than one backup of the full system.
Then if the system does get molested by a preexisting problem
you can step back to the older backup if you need to.

The main advantage of that approach over the backup taken just
after the clean install is that you're likely to have configured stuff
over time since the clean install and those config changes will be
lost if you go back to the backup done just after the clean install.

You're referring to heat - yes?

Cant see what he's on about with 'brackets' unless he is
referring to some drives like the Barracudas that get rid
of heat by conduction to the metal drive bay stack and can get
quite hot if you use plastic rails etc that stop that happening.

I.e., the temperature of HDD's in external enclosures run hotter than when
installed in a professionally built PC?

How about the external enclosures with built in fans? [The ones that actually
have ventilation for moving new air in, hot air out]

They're not usually as well done as the main system case airflow wise.

Lastly, how about 2.5" harddrives? I've got one running in an
aluminum case made by IWILL. It runs cooler to the touch than the
bottom of my notebook where the hard drive is located. Are 2.5"
HDD's any more susceptible to heat induced errors than 3.5" ones?

Nope, if anything less susceptible. They still dont like it tho, the Dell
laptops/notebooks that tend to run the drives right at the max the
drive manufacturer allows are notoriously for being hard drive killers.

 

[cquirke (MVP Windows shell/user)]

So, means you should remove the power cord before removing hard drives?

I would, yes. S-ATA may be hot-swappable if it's using S-ATA power as
well as data connections, but I'd read the small print very carefully
first... in practice, I always pull power out and wait a few seconds.

Right. This 'simply' implies that you need to do a complete system backup
just after installing Windows and setting up your default applications.
Then, keep this known good backup separate from the on-going/periodic
backups, which can be used if you need to restore from hardware failure.

That's the theory, but it's flawed when viewed against 2005 realities.

I might as well write it now... I've been meaning to do a "definitive"
on backup theory for a while, heh.

Backup is inherently a conundrum, if it is to:
- retain all wanted changes (i.e. complete up-to-date contents)
- lose all unwanted changes (i.e. the corruption)

On what basis does one scope in the wanted changes, and scope out the
wanted changes? The usual answer is: Time. Make your backup before
any unwanted changes take place, then you can fall back to this
earlier time at which your PC didn't suck.

Trouble is, any wanted changes done since the backup are lost too.

One of the most effective time-based backups is System Restore, which
is NOT a "backup" in the normal sense of the word. Rather, it is a
system-level "undo" that maintains a series of recent prior states, to
which a system may be returned - as long as the System Restore data is
intact, and the system is still able to do the System Restore.

Your proposal to create a known-good full system backup straight after
the installation (before anything goes wrong) would normally be a fine
way to maintain the system; too bad about any subsequently-installed
programs or subsequently-created data, of course.

So there are two things "wrong" with the above approach:
- it loses all patches, thus leaves you vulnerable to attack
- it loses all apps and data

Nevertheless it's a perfect solution if you are an OEM, responsibile
only for keeping the OS running so that you can refuse an RMA; in
fact, it's the lowest common denominator amongst OEMs, esp. the
biggest brands (where consumer confidence in "the brand" is high, so
they can be gouged and still be likely to come back for more)

Normally, one would modify the approach as "restore the perfect
original boilerplate system backup, re-install any programs installed
since that time, then restore your last data backup".

Implicit in the above are two things:
- an ability to scope data from malicious material
- an assumption that the original "fit to ship" code is fit to use

There's not much grounds for optimism on either count.

Why would you need to restore from backups? The answers to that
question should guide how you try to scope "wanted" from "unwanted"

1) Computer theft, destruction or natural hardware failure
- Wanted: OS and program code, user data
- Unwanted: hardware-specific driver code and settings

2) Malware payload, inability to reclaim ownership from malware
- Wanted: User data, preferences, MRUs
- Unwanted: All potentially-infected code and integrations

3) Software re-versioning
- Wanted: User data, preferences, MRUs, drivers
- Unwanted: Affected code, version dependencies

4) Computer re-deployment, e.g. sale, different job description
- Wanted: OS and program code, drivers
- Unwanted: User data, preferences, settings, MRUs

So in order to prepare for these crises, your management has to be
geared to scoping "wanted" from "unwanted" from the very outset. MS
doesn't really have the faintest clue here; infectable code is mixed
with user data, user data is version-bound to applications that are in
turn version-bound to the OS, even MS Backup is version-bound.

It's a complete mess, made worse by the need to magically be patched
up to date before hitting the Internet for the first time. Frankly,
the best approach is to minimize your data's relationship with MS
products as far as possible, and to override MS's useless duhfault
data locations (e.g. nesting IE and IM downloads and massive Music and
Pictures within the data set, gratuitously long paths, etc.).

You're referring to heat - yes? I.e., the temperature of HDD's in external
enclosures run hotter than when installed in a professionally built PC?

Heat is one of the likely factors, but not the only one, and you can
monitor this as long as the HD is in reach of S.M.A.R.T. queries.

Other factors are; static build-up (ground the HD's shell to chassis),
quality of the data cabling and connections (if you need 80-pin cable,
then an extra set of mechanical contacts will be like hitting a trench
dug across a highway at 70 mph), power misadventures, and mechanics.

How about the external enclosures with built in fans? [The ones that
actually have ventilation for moving new air in, hot air out]

Yes, you'd want non-risible fans, plenty of air circulation, chassis
grounding, decent cabling, good power discipline and gentle handling.
But I've seen sick or dead HDs even when these criteria were met.

Part of the problem may be the choice and quality of HDs adopted by
"removable hard drive" package vendors, such as IOmega etc.

Lastly, how about 2.5" harddrives? I've got one running in an aluminum case
made by IWILL. It runs cooler to the touch than the bottom of my notebook
where the hard drive is located. Are 2.5" HDD's any more susceptible to
heat induced errors than 3.5" ones?

I think 2.5" drives are more failure-prone than 3.5", whether they're
in laptops or in external brackets, even though they are at least
presumably designed to be bumped, powerred up and down, etc.

They are also designed to consume less power, so the performance
usually sucks (think 4500 RPM vs. 7200 RPM, and the puny capacities)
so they should be less likely to heat up.

The pricing's pretty dismal too, but they are small and cute. It
would be interesting to see enough useage data and results to
objectively assess reliability in the following scenarios:
- 3.5" inside PCs
- 3/5" in brackets
- 2.5" inside laptops
- 2/5" in brackets

For example, if 2.5" in brackets fail significantly less often than
3.5" in brackets, then that could make them interesting, even if they
fail in laptops more often than 3.5" fail in "real" PCs.

---------- ----- ---- --- -- - - - -

On the 'net, *everyone* can hear you scream

 

[cquirke (MVP Windows shell/user)]

Nonsense. All the Molex are powered off.
The IDE ports are disabled. Only PS/2, USB, PCI, perhaps RAM are on.
Even disabling IDE channel in Device Manager allows hot swap.

That's an interesting assertion. I can meter molex power leads to
test those, but checking the data cables would be more difficult.

Given the risks involved, I'd like a couple of URLs on that?

------------ ----- ---- --- -- - - - -

The most accurate diagnostic instrument
in medicine is the Retrospectoscope

 

[cquirke (MVP Windows shell/user)]

You've got that backwards. And they can see drives > 137G,
they'll just corrupt the data as access wraps at that point.

Yes indeed; the < should be a >

It would be a lot safer if XP SP0 would simply refuse to "see" an
over-137G than appear to work and then barf the data, so I if
anything, your assertion re-inforces my caveat, unless partitioning
keeps the volumes that are used, short of that point.

Wrong. Have fun explaining xxclone.

I haven't used XXClone, but I've certainly seen what happens when
methods that work flawlessly with Win9x installations are applied to
XP installations. The result doesn't boot, even if you get all the
boot code stuff spot on.

And you dont need a bootable copy for backup anyway.

Depends why you are trying to backup, doesn't it? Someone backing up
the whole of C: is liable to expect a restore to be bootable.

Wrong. And the worst you have to do is unplug
the power supply from the mains anyway.

I'd remove the mains; that's what I'm saying. ATX "off" definately
maintains a standby trickle of power to the system; what may be
debatable is whether this power goes to the HDs, via power leads or
via the data connections.

Bullshit.

Speaking from experience, my friend. I see far more dead HDs from
brackets than I do from "normal" use; why that is, is open to
conjecture, but that's the mileage.

Or are you claiming it's "bullshit" that a malware can permeate a
system, and backups thereof, before hatching a destructive payload
several days later?

Or are you disputing Western Digital's advice NOT to hot-swap their
S-ATA drives if using the legacy power connectors they provide?

Any decent imager allows you to browse the image.

OK, so don't use any of the indecent ones that abound.

-------------------- ----- ---- --- -- - - - -

Reality is that which, when you stop believing
in it, does not go away (PKD)

 

[Rod Speed]

Eric Gisin wrote

That's an interesting assertion.

Its not an assertion, its a fact.

I can meter molex power leads to test those,

And that is all you need to do.

but checking the data cables would be more difficult.

Nope, you can do that with the same multimeter.
At a spare connector on a ribbon cable or the
motherboard end with the ribbon cable unplugged.

Given the risks involved,

There are no risks involved.

I'd like a couple of URLs on that?

Your problem. It should be obvious from what needs
to be powered with an ATX system so that it can be
turned on from the keyboard or wake on lan etc.

 

[Rod Speed]

I would, yes.

No one who understands ATX systems bothers.

S-ATA may be hot-swappable if it's using
S-ATA power as well as data connections,

hot-swappable is an entirely different situation to turning
the system off but not unplugging the mains cord.

but I'd read the small print very carefully first...

You wont get any different answer from reading the SATA standard.

in practice, I always pull power out and wait a few seconds.

Not necessary with a hard drive.

That's the theory, but it's flawed when viewed against 2005 realities.

Nope. Not if you dont mind doing the config again after a restore.

I might as well write it now... I've been meaning to
do a "definitive" on backup theory for a while, heh.

You clearly arent qualified to do that.

Backup is inherently a conundrum, if it is to:
- retain all wanted changes (i.e. complete up-to-date contents)
- lose all unwanted changes (i.e. the corruption)

On what basis does one scope in the wanted changes, and scope
out the wanted changes? The usual answer is: Time. Make your
backup before any unwanted changes take place, then you can
fall back to this earlier time at which your PC didn't suck.

Even that is grossly over simplifying the
story with the system restore available.

Trouble is, any wanted changes done since the backup are lost too.

Depends entirely on how often backups are
done, and what is backed up with each backup.

And what matters most is what is only available
from a backup and what can be done without a
backup which is just a nuisance to have to do.

One of the most effective time-based backups is System Restore,
which is NOT a "backup" in the normal sense of the word.

Corse it is.

Rather, it is a system-level "undo" that maintains a series
of recent prior states, to which a system may be returned

So is a set of backups.

- as long as the System Restore data is intact, and
the system is still able to do the System Restore.

Your proposal to create a known-good full system backup straight
after the installation (before anything goes wrong) would normally
be a fine way to maintain the system; too bad about any subsequently
installed programs or subsequently-created data, of course.

So there are two things "wrong" with the above approach:
- it loses all patches, thus leaves you vulnerable to attack

Not if its done after the patches have been applied.

- it loses all apps and data

It doesnt lose the apps.

And they arent 'lost' in the sense that data
is, they can always be installed again.

Nevertheless it's a perfect solution if you are an OEM, responsibile
only for keeping the OS running so that you can refuse an RMA; in
fact, it's the lowest common denominator amongst OEMs, esp. the
biggest brands (where consumer confidence in "the brand" is high,
so they can be gouged and still be likely to come back for more)

Mindless conspiracy theory. Its all that is that practical for an OEM to do.

Normally, one would modify the approach as "restore the perfect
original boilerplate system backup, re-install any programs installed
since that time, then restore your last data backup".

Implicit in the above are two things:
- an ability to scope data from malicious material

Dont need to if you have a series of backups.

- an assumption that the original "fit to ship" code is fit to use

Tad unlikely that you wouldnt have noticed that it wasnt.

There's not much grounds for optimism on either count.
Bullshit.

Why would you need to restore from backups? The answers to that
question should guide how you try to scope "wanted" from "unwanted"

1) Computer theft, destruction or natural hardware failure

What is unnatural hardware failure ?

- Wanted: OS and program code, user data
- Unwanted: hardware-specific driver code and settings

That last is just plain wrong.

2) Malware payload, inability to reclaim ownership from malware
- Wanted: User data, preferences, MRUs
- Unwanted: All potentially-infected code and integrations

Just a yawn with a series of backups.

3) Software re-versioning
- Wanted: User data, preferences, MRUs, drivers
- Unwanted: Affected code, version dependencies

Completely routine with backups done before the install.

4) Computer re-deployment, e.g. sale, different job description
- Wanted: OS and program code, drivers
- Unwanted: User data, preferences, settings, MRUs

So in order to prepare for these crises,

The last isnt a 'crisis'

your management has to be geared to scoping
"wanted" from "unwanted" from the very outset.

Wrong again.

MS doesn't really have the faintest clue here; infectable code is mixed
with user data, user data is version-bound to applications that are in
turn version-bound to the OS, even MS Backup is version-bound.

You clearly dont have the faintest clue yourself.

It's a complete mess, made worse by the need to magically be
patched up to date before hitting the Internet for the first time.

Wrong again, most obviously with a series of backups.

Frankly, the best approach is to minimize your data's relationship with
MS products as far as possible, and to override MS's useless duhfault
data locations (e.g. nesting IE and IM downloads and massive Music
and Pictures within the data set, gratuitously long paths, etc.).

More mindlessly silly stuff. Its completely trivial to handle
all that effortlessly with any decent approach to backup.

Heat is one of the likely factors, but not the only one, and you can
monitor this as long as the HD is in reach of S.M.A.R.T. queries.

Other factors are; static build-up (ground the HD's shell to chassis),

Happens automatically most of the time.

quality of the data cabling and connections (if you need 80-pin
cable, then an extra set of mechanical contacts will be like
hitting a trench dug across a highway at 70 mph),

Nope, nothing like it.

power misadventures, and mechanics.

Waffle as far as your original silly claim is concerned.

How about the external enclosures with built in fans? [The ones
that actually have ventilation for moving new air in, hot air out]

Yes, you'd want non-risible fans, plenty of air circulation, chassis
grounding, decent cabling, good power discipline and gentle handling.
But I've seen sick or dead HDs even when these criteria were met.

Any mechanical device can die whatever you do.

Just as true of anything used for backup.

Which is why you dont have just one backup.

Part of the problem may be the choice and quality of HDs adopted
by "removable hard drive" package vendors, such as IOmega etc.

Hardly ever.

I think 2.5" drives are more failure-prone than 3.5", whether they're
in laptops or in external brackets, even though they are at least
presumably designed to be bumped, powerred up and down, etc.

You're wrong.

They are also designed to consume less power, so the
performance usually sucks (think 4500 RPM vs. 7200 RPM,

They're still by far the fastest backup devices anyway.

and the puny capacities) so they should be less likely to heat up.

The pricing's pretty dismal too, but they are small and cute.
It would be interesting to see enough useage data and results
to objectively assess reliability in the following scenarios:
- 3.5" inside PCs
- 3/5" in brackets
- 2.5" inside laptops
- 2/5" in brackets

For example, if 2.5" in brackets fail significantly less often than
3.5" in brackets, then that could make them interesting, even if
they fail in laptops more often than 3.5" fail in "real" PCs.

External cases arent 'brackets'

 

[Rod Speed]

Yes indeed; the < should be a >

It would be a lot safer if XP SP0 would simply refuse
to "see" an over-137G than appear to work and then
barf the data, so I if anything, your assertion

It isnt an assertion, its a fact.

re-inforces my caveat, unless partitioning keeps
the volumes that are used, short of that point.

I chose to comment on your claim that the original XP
cant 'see' drives over 137G. That is just plain wrong.

I haven't used XXClone, but I've certainly seen what happens
when methods that work flawlessly with Win9x installations
are applied to XP installations. The result doesn't boot,
even if you get all the boot code stuff spot on.

Irrelevant to your assertion that file level copying isnt possible.

That is just plain wrong. Most of the cloners do it that way.
Essentially because they are mostly used when upgrading
the boot drive to a bigger one and file copying is the most
convenient way to do that.

Depends why you are trying to backup, doesn't it?
Nope.

Someone backing up the whole of C: is
liable to expect a restore to be bootable.

The restore being bootable is an entirely separate
question to whether the backup itself is bootable.

I'd remove the mains; that's what I'm saying.

No need.

ATX "off" definately maintains a
standby trickle of power to the system;
Yes.

what may be debatable is whether this power goes
to the HDs, via power leads or via the data connections.

No its not debateable, its trivial to measure and prove it doesnt happen.

Speaking from experience, my friend.

No you arent, you're actually grossly misusing the term 'brackets'

I see far more dead HDs from brackets than I do from
"normal" use; why that is, is open to conjecture,

Nope, its completely trivial to seen what is different between
an external enclosure and a drive mounted internally.

but that's the mileage.

No it isnt with 'brackets'

Or are you claiming it's "bullshit" that a malware can
permeate a system, and backups thereof, before
hatching a destructive payload several days later?

Nope, that comment applied to just your last line, 6)

Or are you disputing Western Digital's advice NOT to hot-swap their
S-ATA drives if using the legacy power connectors they provide?

Nope, that comment applied to just your last line, 6)

OK, so don't use any of the indecent ones that abound.

The vast bulk of the mainstream imagers allow you to browse the image.

 

[Rod Speed]

Yes indeed; the < should be a >

It would be a lot safer if XP SP0 would simply refuse
to "see" an over-137G than appear to work and then
barf the data, so I if anything, your assertion

It isnt an assertion, its a fact.

re-inforces my caveat, unless partitioning keeps
the volumes that are used, short of that point.

I chose to comment on your claim that the original XP
cant 'see' drives over 137G. That is just plain wrong.

I haven't used XXClone, but I've certainly seen what happens
when methods that work flawlessly with Win9x installations
are applied to XP installations. The result doesn't boot,
even if you get all the boot code stuff spot on.

Irrelevant to your assertion that file level copying isnt possible.

That is just plain wrong. Most of the cloners do it that way.
Essentially because they are mostly used when upgrading
the boot drive to a bigger one and file copying is the most
convenient way to do that.

Depends why you are trying to backup, doesn't it?
Nope.

Someone backing up the whole of C: is
liable to expect a restore to be bootable.

The restore being bootable is an entirely separate
question to whether the backup itself is bootable.

I'd remove the mains; that's what I'm saying.

No need.

ATX "off" definately maintains a
standby trickle of power to the system;
Yes.

what may be debatable is whether this power goes
to the HDs, via power leads or via the data connections.

No its not debateable, its trivial to measure and prove it doesnt happen.

Speaking from experience, my friend.

No you arent, you're actually grossly misusing the term 'brackets'

I see far more dead HDs from brackets than I do from
"normal" use; why that is, is open to conjecture,

Nope, its completely trivial to seen what is different between
an external enclosure and a drive mounted internally.

but that's the mileage.

No it isnt with 'brackets'

Or are you claiming it's "bullshit" that a malware can
permeate a system, and backups thereof, before
hatching a destructive payload several days later?

Nope, that comment applied to just your last line, 6)

Or are you disputing Western Digital's advice NOT to hot-swap their
S-ATA drives if using the legacy power connectors they provide?

Nope, that comment applied to just your last line, 6)

OK, so don't use any of the indecent ones that abound.

The vast bulk of the mainstream imagers allow you to browse the image.

 

[Richard Urban [MVP]]

I would think you would like to back up what you state with a few quality
url's as proof. Otherwise, you are blowing smoke.

--
Regards,

Richard Urban
Microsoft MVP Windows Shell/User

Quote from: George Ankner
"If you knew as much as you think you know,
You would realize that you don't know what you thought you knew!"

 

[Rod Speed]

I would think

Not a shred of evidence that you are actually capable of thought.

You clearly couldnt bullshit your way out of a wet paper
bag even if your pathetic excuse for a 'life' depended on it.

you would like to back up what you state with a few quality url's as proof.

Dont need any of those with a question as basic as the one being discussed.
Even someone as stupid as you should be able to check whether hard drives
are powered up when an ATX system is shutdown, using such simple approachs
as checking if there is any power on the molex connector and it only requires
a multimeter and something viable between your ears to check the data cable.

Otherwise, you are blowing smoke.

See above on the wet paper bag.

 

[Ge]

I'd remove the mains; that's what I'm saying. ATX "off" definately
maintains a standby trickle of power to the system;

No, not really. No digital system works with a "trickle of power". It's
either enough power to supply what's connected to it properly, or it isn't.
If it isn't, the power supply is not adequate for the system, and the
system may function partially, but won't function properly. I guess we can
assume that we are talking about a PC that does function properly.

(There is actually a term "trickle supply", but that's more for battery
chargers and similar circuits; it usually refers to the small current
that's necessary to keep a charged battery fully charged. The ATX standby
supply is not what's commonly called a "trickle supply"; it is a full-blown
+5 V supply that supplies at least 1 A for standby power -- just as its
name says.)


The ATX standby power is only provided on the ATX standby power line
(called +5VSB); all other power lines are "off" in standby mode. There's no
"trickle of power": the +5VSB line is fully on, all others are fully off.

What is in fact connected to the +5VSB (ATX standby power) is dependent on
the specific mobo and the BIOS configuration, of course, but with the ATX
standard guaranteeing no more than 1 A on that line, and it being only +5 V
(no +12 V standby power), I doubt there are any decent mobos out there that
supply their IDE controllers with standby power. If they did, that would be
a serious design flaw, given that the harddisk that's connected to that
connector in any case is not supplied with power during standby (because
the standard harddisk power connector doesn't carry the +5VSB standby power
line).

While it is theoretically thinkable that a ill-designed mobo supplies its
IDE controlles from ATX standby power, it is still not clear whether this
does any harm to the harddisk. If it does, it probably does so whether or
not you disconnect the harddisk while power is supplied to its IDE
interface -- the damage being done by it being connected, not by the act of
connecting or disconnecting it.

Since there have been questions about "serious links" here... how about
some serious links confirming that there are indeed mobos that do supply
their IDE controllers with standby power (which I doubt)? Or links about
how harddisk controllers deal with their IDE interface being powered while
their own power supply is off (probably a situation deemed out of specs by
the drive manufacturer, therefore any mobo that would do that is not
adequate anyway)? Regarding links to back up what I said... that's all just
explaining the ATX standard. You should be able to find any number of links
to information about this.


OTOH, telling a newbie to disconnect the power cord before opening the PC
case can be good advice. But not because of "power trickles" in the
computer... This good advice should come with an equally good explanation,
like "As long as you're not knowing what you're doing, it's a good idea to
always disconnect the power cord of /anything/ you open."

Gerhard

 

[J. Clarke]

No, not really. No digital system works with a "trickle of power". It's
either enough power to supply what's connected to it properly, or it
isn't. If it isn't, the power supply is not adequate for the system, and
the system may function partially, but won't function properly. I guess we
can assume that we are talking about a PC that does function properly.

(There is actually a term "trickle supply", but that's more for battery
chargers and similar circuits; it usually refers to the small current
that's necessary to keep a charged battery fully charged. The ATX standby
supply is not what's commonly called a "trickle supply"; it is a
full-blown +5 V supply that supplies at least 1 A for standby power --
just as its name says.)


The ATX standby power is only provided on the ATX standby power line
(called +5VSB); all other power lines are "off" in standby mode. There's
no "trickle of power": the +5VSB line is fully on, all others are fully
off.

What is in fact connected to the +5VSB (ATX standby power) is dependent on
the specific mobo and the BIOS configuration, of course, but with the ATX
standard guaranteeing no more than 1 A on that line, and it being only +5
V (no +12 V standby power), I doubt there are any decent mobos out there
that supply their IDE controllers with standby power. If they did, that
would be a serious design flaw, given that the harddisk that's connected
to that connector in any case is not supplied with power during standby
(because the standard harddisk power connector doesn't carry the +5VSB
standby power line).

While it is theoretically thinkable that a ill-designed mobo supplies its
IDE controlles from ATX standby power, it is still not clear whether this
does any harm to the harddisk. If it does, it probably does so whether or
not you disconnect the harddisk while power is supplied to its IDE
interface -- the damage being done by it being connected, not by the act
of connecting or disconnecting it.

Since there have been questions about "serious links" here... how about
some serious links confirming that there are indeed mobos that do supply
their IDE controllers with standby power (which I doubt)? Or links about
how harddisk controllers deal with their IDE interface being powered while
their own power supply is off (probably a situation deemed out of specs by
the drive manufacturer, therefore any mobo that would do that is not
adequate anyway)? Regarding links to back up what I said... that's all
just explaining the ATX standard. You should be able to find any number of
links to information about this.


OTOH, telling a newbie to disconnect the power cord before opening the PC
case can be good advice. But not because of "power trickles" in the
computer... This good advice should come with an equally good explanation,
like "As long as you're not knowing what you're doing, it's a good idea to
always disconnect the power cord of /anything/ you open."

You're being pedantic here IMO. +5VSB is a "trickle" compared to the normal
operating power. It seems a pointless argument that you are raising.

The point is that there is power in the case and on the motherboard with the
ATX power in the "off" state. And depending on what you're doing that
standby power can become full power very easily and very quickly--I've
inserted boards into or removed boards from a machine and had it power up
while I was doing it (and don't tell me it can't happen based on some
fancy-Dan analytical model--in the real world it can and does), sometimes
to the detriment of one component or another.

When adding or removing hardware, the best policy is a cold chassis, that
means either unplugged or hard-off.

 

[cquirke (MVP Windows shell/user)]

Its not an assertion, its a fact.

It's not a fact because you say it is. Without proof, it's an
assertion, and one that I happen to be interested in.

But your attitude is making me less interested by the minute.

And that is all you need to do.

Nope. More on that later.

Nope, you can do that with the same multimeter.
At a spare connector on a ribbon cable or the
motherboard end with the ribbon cable unplugged.

Quite a lot of combinations to test, even when you discount 40 of the
80 lines as common ground. Also, meters are fairly crude, when it
comes to transients - if a data line was to pulse once every 1/10th of
a second, it may not show up on a meter at all.

There are no risks involved.

Apart from crucial data on a costly hard drive, you mean?

Your problem. It should be obvious from what needs
to be powered with an ATX system so that it can be
turned on from the keyboard or wake on lan etc.

No, it's not my problem - I can simply discard your assertions as
unsubstantiated and move on, if you can't back them up. As long as
there's a risk of power being applied to devices while "switched off",
I'll advise not plugging them in or out unless you are certain it's
safe to do so. You may have that certainty; I do not.

------------------------ ---- --- -- - - - -

Forget http://cquirke.blogspot.com and check out a
better one at http://topicdrift.blogspot.com instead!

 

[cquirke (MVP Windows shell/user)]

On 8/14/05 19:43:29, cquirke (MVP Windows shell/user) wrote:

No, not really. No digital system works with a "trickle of power". It's
either enough power to supply what's connected to it properly, or it isn't.

I was referring to the standby power line that you mention a bit
further on. On that line, it's a regular 5V; when looking at the
total current draw from the PSU, it appears as a "trickle" there.

Still, enough to get into trouble. I saw a particular line of power
supplies that would tend to heat up when ATX "off" (the top of the
case would be warmer than when the PC was running), and sometimes
these would go Bang! and blow a cap when switched on.

This would then spike the +12V and typically kill motherboard, RAM,
HD, and the CD-ROM would often fail a week or so later.

One possible mechanism for this failure pattern might be that the
lower standby current with no fan caused the PSU to "run" hotter than
normal operations with the fan running. Then when the initial demands
of full power are applied to this heated circuitry, pop!

Or it could be some other mechanism entirely, unrelated to the warmer
temperature noted when "just" supplying standby power.

The ATX standby power is only provided on the ATX standby power line
(called +5VSB); all other power lines are "off" in standby mode. There's no
"trickle of power": the +5VSB line is fully on, all others are fully off.

So what we need to know is where that line goes., and whether power
from that line emerges through other ciruitry as data line activity.

What is in fact connected to the +5VSB (ATX standby power) is dependent on
the specific mobo and the BIOS configuration, of course, but with the ATX
standard guaranteeing no more than 1 A on that line, and it being only +5 V
(no +12 V standby power), I doubt there are any decent mobos out there that
supply their IDE controllers with standby power. If they did, that would be
a serious design flaw, given that the harddisk that's connected to that
connector in any case is not supplied with power during standby (because
the standard harddisk power connector doesn't carry the +5VSB standby power
line).

Well, it could be some sort of "ghost power" arrangement I suppose,
much as some USBs which have external power connections "don't need
external power" and can run without it.

I notice some mobos support wake-up from various IRQs, including the
IDE controllers (wake-on-demand is not a not a feature I use, so I
don't spend much time in that part of CMOS setup). If the mobo can
sense events on the IDE, then that suggests power on the IDE.

While it is theoretically thinkable that a ill-designed mobo supplies its
IDE controlles from ATX standby power, it is still not clear whether this
does any harm to the harddisk. If it does, it probably does so whether or
not you disconnect the harddisk while power is supplied to its IDE
interface -- the damage being done by it being connected, not by the act of
connecting or disconnecting it.

Again, maybe and maybe not. It's quite possible to circuit something
that's safe for isolated power, i.e. a logic chip that derives power
from one of the IDE lines and doesn't rely on other HD components that
would have to formally powered via the molex.

Since there have been questions about "serious links" here... how about
some serious links confirming that there are indeed mobos that do supply
their IDE controllers with standby power (which I doubt)? Or links about
how harddisk controllers deal with their IDE interface being powered while
their own power supply is off (probably a situation deemed out of specs by
the drive manufacturer, therefore any mobo that would do that is not
adequate anyway)? Regarding links to back up what I said... that's all just
explaining the ATX standard. You should be able to find any number of links
to information about this.

I don't have links, but I reckon if there's doubt on this, it's safer
not to assume it's safe to plug and unplug HDs while the ATX power is
connected but "switched off". We do know that while "off", power
enters the PC as a whole, and the mobo too; we are debating whether it
gets to the HDs. Until certain it's safe, I'd rather unplug the power

OTOH, telling a newbie to disconnect the power cord before opening the PC
case can be good advice. But not because of "power trickles" in the
computer... This good advice should come with an equally good explanation,
like "As long as you're not knowing what you're doing, it's a good idea to
always disconnect the power cord of /anything/ you open."

Yep. In the old days of "real" power switches, there was a school of
thought that suggested it was safer for the PC's parts if mains was
plugged in but switched off, in that the system would still be
grounded via the earth line of the mains supply.

I'd say that's changed now, in the ATX era; if it ever was good
advice, it's less so these days.

---------- ----- ---- --- -- - - - -

Proverbs Unscrolled #37
"Build it and they will come and break it"

 

[Arno Wagner]

On 8/14/05 19:43:29, cquirke (MVP Windows shell/user) wrote:

[ correct explanaition about "trickle" and +5VSB snipped]

While it is theoretically thinkable that a ill-designed mobo supplies its
IDE controlles from ATX standby power, it is still not clear whether this
does any harm to the harddisk. If it does, it probably does so whether or
not you disconnect the harddisk while power is supplied to its IDE
interface -- the damage being done by it being connected, not by the act of
connecting or disconnecting it.

I can comment on this:

1. +5VSB will never be supplied to the HDDs, since it does not have
enough current reserves.

2. If there was 5V applied to a HDD and the 12V line was grounded
(even over other devices) or set to 12V the HDD electronics could
be destroyed (thing smoke and possibly exploding components) if
the +12V and +5V lines make contact before any of the ground
lines make contact. The SATA power connector is designed
specifically to avoid this. The standard molex connector is
not.

3. Applying +5V but not +12V to a HDD should not damage it in any way.
The HDDs monitor their power lines (needed for save shutdown) and
will know that +12V is missing or out of spec. Also some resiliance
against one or both voltages being too low is needed anyways, since
this situation will be present for a short time during start-up and
shut-down. In addition there is no technical reason to design
a circuit to be vulnerable to this condition and it is very easy to
design it to be not vulnerable.

4. Logic-lines (IDE interface) are usually permanently short circuit
proof today and the other side (the unpowered one) is usually
resilient to input currents up to the maximum short-circuit
current. I.e. connectin a powerd logic chip output (input) to
an input (output) of an unpowerd logic chip usually does not
do damage. Also if HDDs were damaged frequently when somebody
forgot to plug in power when installing a new HDD, this would
be knowen by now.

Since modern HDDs have controller-integrated bus drivers, I cannot
provide a link, but looking into the specification of any ATAPI
usable line-buffer chip should tell you this. An example would
be the 74ALS540, which can withstand up to 7V input voltage,
regardless of its positive supply voltage.

To sum up: Connecting or disconnectin HDD power when voltages are
present is extremely risky with anything except SATA power
connectors. Having the IDE cable plugged in before (after on removal)
_may_ provide the needed ground connection to the HDD, but don't depend
on it. On the other hand, connecting or disconnecting the IDE cable
should be a minor or no risk even if power is present on one or both
sides.

Since there have been questions about "serious links" here... how about
some serious links confirming that there are indeed mobos that do supply
their IDE controllers with standby power (which I doubt)?

Since this would be complete nonsense, I doubt there is any link.
Why spend the extra money to give the controller a power-switch
that switches the controller over between +5VSB and +5V, when the
controller is perfectly happy to work on +V5?

Or links about
how harddisk controllers deal with their IDE interface being powered while
their own power supply is off (probably a situation deemed out of specs by
the drive manufacturer, therefore any mobo that would do that is not
adequate anyway)?

Sadly there seems to be no information by drive manufaturers or
in the ATAPI standards about this. So only the general reasoning
about modern logic bus drivers above applies.

Regarding links to back up what I said... that's all just
explaining the ATX standard. You should be able to find any number of links
to information about this.

OTOH, telling a newbie to disconnect the power cord before opening the PC
case can be good advice. But not because of "power trickles" in the
computer... This good advice should come with an equally good explanation,
like "As long as you're not knowing what you're doing, it's a good idea to
always disconnect the power cord of /anything/ you open."

Actually it is good advice because +5VSB does go to some devices like
the PCI cards. And the current available can be enough to do damage
if shortened out.

Arno

 

[Rod Speed]

Rod Speed wrote

It's not a fact because you say it is.

Its a fact regardless of what I say.

Without proof, it's an assertion,

Wrong again, its still a fact. I have just stated that its a fact.

And I have told you how to prove that fact, its completely trivial
to check if the hard drives are powered when the system is shutdown
but still has the mains cord connected to the ATX power supply.

and one that I happen to be interested in.

You have always been, and always will be, completely and utterly irrelevant.

But your attitude is making me less interested by the minute.

You have always been, and always will be, completely and utterly
irrelevant. What you may or may not claim to be interested in in spades.

Yep.

More on that later.

Completely useless.

Quite a lot of combinations to test, even when you
discount 40 of the 80 lines as common ground.

Still completely trivial to do, even you should be able
to manage it. And you dont measure 'combinations',
you measure the voltage between ground and each pin.

Also, meters are fairly crude, when it comes to transients

Irrelevant. There wont be any 'transients' with the system shut down.

- if a data line was to pulse once every 1/10th of
a second, it may not show up on a meter at all.

Wrong again. You'll certainly see something, you just wont
be able to measure the pulse accurately. In fact you wont
see anything because there wont be any pulses on those lines.

And while someone as ignorant as you wont be
able to use a CRO, or have one either, thats easy
for anyone with a clue to check that fact with too.

Apart from crucial data on a costly hard drive, you mean?

Measuring if there is anything on the ribbon cable with it unplugged
from the drive doesnt risk the data on the drive at all.

Ditto with measuring the voltages on unused molex connectors too.

No, it's not my problem

Corse it is.

I can simply discard your assertions as unsubstantiated
and move on, if you can't back them up.

I dont need to 'back them up' when even someone as stupid as
you should be able to do the measurements above if you doubt it.

You arent prepared to do that ? YOUR problem.

As long as there's a risk of power being
applied to devices while "switched off",

There is no risk whatever of that with hard drives.

I'll advise not plugging them in or out
unless you are certain it's safe to do so.

And we can all laugh at your pig ignorance,
and bone headed stupidity too.

Its actually better to leave the mains cord plugged into the power
supply, so the system remains grounded thru the power cord.

You may have that certainty; I do not.

Your problem, as always.

 

[TonyB]

You're being pedantic here IMO.
Nope.

+5VSB is a "trickle" compared to the normal operating power.

YOU are being pedantic in fact.

It seems a pointless argument that you are raising.

Corse it isnt when discussing whether its safe to unplug a hard
drive with an ATX system shutdown and the mains still connected.

The point is that there is power in the case and on the motherboard
with the ATX power in the "off" state. And depending on what you're
doing that standby power can become full power very easily and very
quickly--I've inserted boards into or removed boards from a machine
and had it power up while I was doing it (and don't tell me it can't
happen based on some fancy-Dan analytical model--in the real world it
can and does),

How odd that its never happened to me once.

sometimes to the detriment of one component or another.

When adding or removing hardware, the best policy is
a cold chassis, that means either unplugged or hard-off.

Not with just changing a hard drive it isnt. THAT is what was being discussed.