Changing PCI cards without switching off?

  • Thread starter Thread starter Franklin
  • Start date Start date
Alex Fraser responded:
But the possibility of damage is a moot point: every time I've knocked a
loose PCI card (physical shock to the PC, or catching some protruding
connector/attached cable), the PC has crashed.
Click to expand...

That's useful to know.
If removing a card is
practically guaranteed to crash the PC (and it would seem so), then you
might as well turn it off anyway.
Click to expand...

Yes -- I don't actually advocate hot-swapping standard PCI cards, but
was interested to find if there was any actual information rather than
just (eminently sensible) warnings about the consequences. I note that
nobody has yet reported damage due specifically to hot-plugging PCI
cards. If I remember and have time, I might do some tests the next time
I scrap a working machine.

By the way, even if this sort of thing doesn't seem to do damage, it's
not a good idea without longer-term testing to ensure that circuitry
doesn't suffer non-fatal damage leading to unreliable operation.

I was prompted to ask this question remembering a book (The Birth of a
New Machine?) about a computer developed under time pressure by DEC (?).
What struck me was that they regularly hot-swapped just about any
socketed chip (which is always considered to be something you should
absolutely never do) with no trouble. Computers had a lot of chips in
those days, rather than the few large-scale integrated chips used now.

Best wishes,
 
There are a lot of people warning that removing a PCI card from a
running machine will cause damage. While I don't intend to do this, has
anybody actually tested removing and inserting PCI cards in running
machines, or is this just the Standard Warning? (Easy enough to test
these days, with older machine being scrapped because you can't give
them away.) I would hazard a guess that it might actually not cause
damage in most, or all, cases.
[/QUOTE]
Actually it can toast the PCI controller and any other car on the PCI
bus.
Certainly there are some flavours of PCI designed for hot-swapping.
Click to expand...
And you know you've got them because you'll certainly pay for them.
 
Conor wrote about hot-swapping PCI cards:
Actually it can toast the PCI controller and any other card on the PCI
bus.
Click to expand...

Have you ever known a case where this has actually happened? I'm not
trying to say that it can't happen, merely trying to find specific
instances rather than warnings not based on actual experience.

Even if nobody has known it to cause damage, I don't advocate the
practice.
 
Michael said:
Conor wrote about hot-swapping PCI cards:


Have you ever known a case where this has actually happened? I'm not
trying to say that it can't happen, merely trying to find specific
instances rather than warnings not based on actual experience.
Click to expand...

It is just common sense. If the lines are connected in no particular
order, then you could possibly connect the positive pole of the power
and a signal line before the negative power pole was connected. That
could then easily fry circuits between the signal and the positive pole.
Even if nobody has known it to cause damage, I don't advocate the
practice.
Click to expand...

Don't even think about it...
 
No one has provided the proper technical answer. First
removing. Damage most easily caused when a card uses multiple
voltages. Simply breaking a ground connection before other
voltages then causes excessive or reversed voltages on some PN
(semiconductor) junctions is excess of spec. IOW it destroys
transistors.

Busses are designed to be 'hot popped'. There is even a PCI
standard for hot popping. First and foremost, every card must
be designed so the ground pin is always 'last to break and
first to make'. Then other complications only make things
worse.

Now for connecting. Same problem exists for power on.
Suppose the 5 volts arrives before 12 volts. Now junctions
that saw +7 volts (difference between 12 and 5 volt busses)
are instead powered by -5 volts. Not good. Furthermore,
there is a slight 'power on' current surge. It does not exist
during normal 'power on' because power cycling the power
supply causes such a slow and gentle voltage change. But 'hot
popping' a card into a hot buss means a sudden jolt of
current. This can cause data errors (also known as computer
crash) and may even cause hardware damage to PCI pins not
rated for so much current or performing as a switch. Devices
that switch power on and off must be specially designed to do
so. PCI pins are not designed for switching. Numerous
reasons why cards that are hot popped also include current
limiter circuits for hot popping.

These are only a sample in a long line of reasons that make
'hot popping' designs so complicated. Learn about the 'hot
pop' versions of PCI. It is a kludge solution, but it
demonstrates omplexity to avoid hardware damage. Ssolid
technical reasons demonstrate why 'hot popping' is hardware
destructive.
 
Johannes said:
It is just common sense. If the lines are connected in no particular
order, then you could possibly connect the positive pole of the power
and a signal line before the negative power pole was connected. That
could then easily fry circuits between the signal and the positive pole.
Click to expand...

Everybody but everybody just knows that it will cause damage (for
perfectly good technical reasons) but nobody at all seems to have come
across an instance where damage has actually happened!

I have seen plenty of things happen that are supposed to cause damage,
and haven't. I would expect (but haven't checked) that the pinout of PCI
devices is designed so that it is likely that contacts are made and
broken in a safe order when cards are plugged in and out. Also, the
circuitry and components used may be more forgiving than they could be.

Things that, to my surprise, did not cause damage include removing
memory DIMMs from a powered-up (oops) motherboard; connecting an 18V
"wall-wart" PSU to a 5V router (I don't remember if polarity was
correct, inverted, or ac); inserting connectors and chips the wrong way
round. And of course, routinely inserting and removing DIL chips from a
running computer, as done by a company developing a new computer in the
days of DTL and TTL (as mentioned in an earlier posting of mine).

Another thing that I would expect to cause damage is random shorts to
earth; for example, via mounting pillars on an incorrectly mounted
motherboard. But no, remove the short and it all springs into life.

While bad practice shouldn't be encouraged, you often get away with
mistakes.

Best wishes,
 
....
Yes -- I don't actually advocate hot-swapping standard PCI cards, but
was interested to find if there was any actual information rather than
just (eminently sensible) warnings about the consequences. I note that
nobody has yet reported damage due specifically to hot-plugging PCI
cards.
Click to expand...

I read that some have.

If you want electrical/electronic explanations, maybe you should post
to (sci.electronics...) groups.
 
Michael Salem said:
Everybody but everybody just knows that it will cause damage (for
perfectly good technical reasons) but nobody at all seems to have
come across an instance where damage has actually happened!
Click to expand...

That makes at least the third time you have said that in this
thread. And that disagrees with what I have read. Maybe your news
server is incomplete?

Again, if you want electrical/electronic information, you can get
that in a (sci.electronics...) discussion group. When you do, please
feel free to post your enlightenment back here.

Good luck.
 
Michael said:
[...]

While bad practice shouldn't be encouraged, you often get away with
mistakes.
Click to expand...

And sometimes not so lucky. As badly made 3.5" -> 2.5" laptop disk
drive connector once caught the chassis and short circuited the power
wires. The sudden pulse fried the circuit board on my laptop drive!
 
w_tom said:
No one has provided the proper technical answer.
Click to expand...

The original cross posted question has been answered just fine.
Busses are designed to be 'hot popped'.
Click to expand...

http://tinyurl.com/666mp

(http://groups-beta.google.com/groups?as_epq=hot popped&saf
e=images&as_ugroup=Sci.electronics*&as_scoring=d&lr=lang_en&
num=100&hl=en)

There is no such thing as "hot popped".

Once again, w_tom helps prove why posting off-topic can lead to
problems.

w_tom, please take your engineering degree to a (sci.electronics...)
group. I enjoy the reading.
















There is even a PCI
 
Michael Salem said:
There are a lot of people warning that removing a PCI card from a
running machine will cause damage. While I don't intend to do this, has
anybody actually tested removing and inserting PCI cards in running
machines, or is this just the Standard Warning? (Easy enough to test
these days, with older machine being scrapped because you can't give
them away.) I would hazard a guess that it might actually not cause
damage in most, or all, cases.
Click to expand...

Heck, I've pulled BIOS chips out of a running PC. There's a good chance
something will be damaged when disconnecting a card or chip from a powered
up (plugged in!) PC.

Just DON'T do it unless you have no other alternative!!!
 
Michael Salem said:
Alex Fraser responded:
Click to expand...

I probably should have added that this hasn't happened a huge number of
times, perhaps a dozen or so, but other people have had identical
experiences. Just to be clear, the PC crashing is the only apparent effect.
By the way, even if this sort of thing doesn't seem to do damage, it's
not a good idea without longer-term testing to ensure that circuitry
doesn't suffer non-fatal damage leading to unreliable operation.
Click to expand...

Indeed, that or some kind of degradation that ultimately leads to failure.

Alex
 
Everybody but everybody just knows that it will cause damage (for
perfectly good technical reasons) but nobody at all seems to have come
across an instance where damage has actually happened!
Click to expand...
Seeing as I ****ed an IDE controller up once pulling the ribbon cable
off a running drive then a PCI slot that has more power running down it
is more likely to.

Also, there's a reason why you are told to turn off the power first.
I have seen plenty of things happen that are supposed to cause damage,
and haven't. I would expect (but haven't checked) that the pinout of PCI
devices is designed so that it is likely that contacts are made and
broken in a safe order when cards are plugged in and out. Also, the
circuitry and components used may be more forgiving than they could be.
Click to expand...
What order they're made is irrelevent. At the end of the day there's
going to be a power spike when you put it back in. Hotswap PCI gear is
buffered against that.
 
hot swapped a modem once [was stressed, rushing - was working at a firm and
they really piled on the pressure]

hot swapped an ambient modem for a connexant.

Realised my mistake, rebooted the system, the system detected the new modem
and worked fine :-)

The ambient modem worked fine in another system too.

I wouldnt recommend trying it though....


--
No66y©
Those who find they're touched by madness
Sit down next to me

Reply to address is a spam trap.
Use no66y [at] breathe [dot] com
 
Doing something is just fine only because it did not cause
failures previously? What kind of logic is that? Some
darkening of the first O'ring was the only apparent effect
when launching a space shuttle in sub freezing weather. That
proved, beyond a doubt to every bean counter type, that the
shuttle could launch in sub freezing weather. At least until
Challenger exploded. Then Morton Thiokol almost got away with
the coverup. Every engineer said don't launch. They understood
one must first know why it does not cause failure. The 'why'
says PCI card damage can be created by hot popping - and not
just system crashes.

Yes hot popping peripheral cards will not always cause
hardware damage. Damage is downright infrequent since many IO
cards do not use multiple voltages and draw very little
current. But explained in technical detail are examples of
how a PCI card can be damaged AND how I suspect one may have
been damaged. It is rare, but it does happen when it should.

Why does a PCI standard for a unique PCI bus exist so that
cards can be hot popped? So that system hardware could be
reconfigured without interrupting system operation. A special
PCI bus that requires additional circuitry and LEDs on the
motherboard. Point is that a special standard exists to avoid
PCI hardware damage because standard PCI buss cannot be 'hot
popped'. The basic PCI bus does not permit hot popping. Even
if damage does not immediately occur, over stress may cause
PCI card failure later. Posted previously was why hot popping
causes PCI damage to happen and the conditions that make this
failure possible.

Even the USB standard has special hardware and provisions so
that USB devices can be 'hot popped'. Connecting and
disconnecting devices from a buss requires special hardware so
that hot popping causes neither system crashes nor hardware
damage. System crashing alone is not the only symptom.
 
My dear sir: Do you ever first learn what you are reading
before posting like an business school graduate? Shame on
you. Cited in
http://tinyurl.com/666mp
are discussions about voltages tripping out. For example:
I did have 25"'ers pop their HOTs because of that L14401. Even
my CTC185 19" set came in HOT popped because of that infamous
"L14401".
Click to expand...
Why are voltages in a monitor relevant to hot popping PCI
peripheral cards? Shame on you for posting only because
another discussion used the same phrase. You are an expert
only because you can do Google searches? First you must
comprehend the science behind that Google text. Your post
demonstrates why a newsgroup poster again earns bad
credibility. News flash - hot popping is permitted by some
busses and not in others. Hot popping exists.

Properly posted with appropriate technical reasons are why
PCI cards cannot be hot popped. A post that every technically
knowledgeable computer assembler would understand and
appreciate. Unfortunately it is too complex for John Doe who
wants it posted to other newsgroups.

Only the naive John Doe would believe hot popping does not
exist because he could not find it in Google. As Bugs Bunny
once said, "What a Maroon".
 
Rene said:
There is one reason why hardware interrupts were invented, and that is so
that polling wouldn't be necessary anymore.
But apart from that, You are very right saying it would be very unwise to do
what OP wants to do.

Greetings,
Rene
Click to expand...

I don't want to kick a dead horse, but I must ask you a question: Are
you saying then that assuming no damage was done in removing and
re-installing the PCI modem with power on, that the system would still
think it had the same modem as before it was removed? I have not pulled
many such cards, but I seem to recall that even disconnecting a HD from
the IDE bus caused the drive to be removed from the hardware it thought
it had. Needless to say I did not do this often enough to be certain of
the outcome. I look forward to your reply.
 
"w_tom" wrote in message
Even the USB standard has special hardware and provisions so
that USB devices can be 'hot popped'. Connecting and
disconnecting devices from a buss requires special hardware so
that hot popping causes neither system crashes nor hardware
damage. System crashing alone is not the only symptom.
Click to expand...

I read that and immidiately the words "Bill Gates" + "USB" + "press
conference" +" Blue screen of death" +" embarassed looks" spring to
mind.....

:-)


--
No66y©
Those who find they're touched by madness
Sit down next to me

Reply to address is a spam trap.
Use no66y [at] breathe [dot] com
 
Back
Top