Can ATX PSU blow the mainboard?

[Jason]

Can an overheating ATX PSU blow the mainboard if the PSU fails?

I may be mistaken but ISTR there's a possibility of a voltage surge when
the PSU goes.

 

[Jaimie Vandenbergh]

Can an overheating ATX PSU blow the mainboard if the PSU fails?

I may be mistaken but ISTR there's a possibility of a voltage surge when
the PSU goes.

A good one should not. But then even a good one that's been overheated
for too long may no longer be any good. And a bad one could do
anything.

So it's certainly possible. Get a new PSU and see if the machine comes
back up. If not, it's time to spend more money.

Cheers - Jaimie

 

[Johny B Good]

Can an overheating ATX PSU blow the mainboard if the PSU fails?

I may be mistaken but ISTR there's a possibility of a voltage surge when
the PSU goes.

Apart from the well known infamous Bestec ATX psus used by eMachines
models harking back a decade ago, most ATX PSUs fail without damaging
the MoBo.

The usual failure mode when SMPSUs overheat is for the switching
transistors in the HT module to go short circuit and smash the safety
fuse to smithereens in a flash of light accompanied by a loud pop and
a trace of smoke.

When a PSU shuffles off its mortal coil with such drama, only a
single modest transient voltage spike appears on the voltage rails
(the very small ferrite transformer can only transfer a limited amount
of power due to such a _single_ terminal pulse of current) which is
incapable of causing damage. IOW, the kindest failure mode wrt the
load is that dramatic sudden fuse shattering failure of the switching
transistors in the HT module.

There are other types of failure involving sustained overvoltage but
these tend to be rare unless the manufacturer goes out of his way to
bypass/ignore the overload/overvoltage protection features built into
all the standard smpsu controller chips that made it possible to
manufacture ATX psus at all in large volumes at affordable prices.

If the ATX PSU in question went "BANG!", then it's most likely that
no harm has been done and a drop in replacement ATX PSU will get
things working again.

 

[Philip Herlihy]

.... (good stuff) trimmed.

If the ATX PSU in question went "BANG!", then it's most likely that
no harm has been done and a drop in replacement ATX PSU will get
things working again.

An acquaintance brought me a machine which had gone bang in this way
last year. I dropped in a cheap replacement PSU, and it seemed fine. I
left it running overnight. In the morning it was inert. Tested the PSU
- now dead. Tested a second PSU and dropped that in. Still inert.
Tested the second replacement PSU again - dead. This problem was
clearly out of my pay-grade, so sent it to a normally reliable repairer
I'd used before. They said they couldn't fix it - motherboard had gone
- and advised sending it back to the manufacturer (Carillon). They
replaced the motherboard, and also the hard disk, discarding the old
one, along with all my acquaintance's most precious data. Ex-
acquaintance now - he blamed me, and was very unpleasant. I guess I
could have backed up his data for him, but he hadn't, and wanted
everything on the cheap. People often won't pay for the time it takes
to do things properly, and then there's outrage when something goes
wrong (thankfully very rare so far).

Oddly, I've just told a valued customer (valued mainly because he's such
a nice bloke) that I'll do his repair for free now because it's taken so
long to find time to do it. And I realise I'm immediately taking more
precautions (e.g. imaging everything plus data snapshots) because it
won't cost him any more. I'm not cut out for this work...

 

[Paul]

An acquaintance brought me a machine which had gone bang in this way
last year. I dropped in a cheap replacement PSU, and it seemed fine. I
left it running overnight. In the morning it was inert. Tested the PSU
- now dead. Tested a second PSU and dropped that in. Still inert.
Tested the second replacement PSU again - dead. This problem was
clearly out of my pay-grade, so sent it to a normally reliable repairer
I'd used before. They said they couldn't fix it - motherboard had gone
- and advised sending it back to the manufacturer (Carillon). They
replaced the motherboard, and also the hard disk, discarding the old
one, along with all my acquaintance's most precious data. Ex-
acquaintance now - he blamed me, and was very unpleasant. I guess I
could have backed up his data for him, but he hadn't, and wanted
everything on the cheap. People often won't pay for the time it takes
to do things properly, and then there's outrage when something goes
wrong (thankfully very rare so far).

Oddly, I've just told a valued customer (valued mainly because he's such
a nice bloke) that I'll do his repair for free now because it's taken so
long to find time to do it. And I realise I'm immediately taking more
precautions (e.g. imaging everything plus data snapshots) because it
won't cost him any more. I'm not cut out for this work...

If you plan to repair a lot of PCs, I recommend a couple things.

1) Everyone should have a multimeter, because you can
get a basic one for $20. And it allows checking
the voltages. So bang for buck, it's a good deal.

Measuring current flow, requires a breakout
cable of some sort - you need to be "in series"
when using one of those, to measure current flow.
Measuring all the rails would take all morning.

2) A second item, more expensive, is a DC clamp-on ammeter.
I can wrap the jaws of that, around wires of the same color
on the main ATX power cable, and make current measurements.
And see whether the load is normal or not. Or whether
the load is really too much for the replacement supply.
There are many AC clamp-on ammeters, but DC capable ones
are fewer in number. The DC ones use a Hall probe for
the measurement.

I got one like this, an Extech, before they got the ugly
color scheme. Mine is grey in color, throughout. For PC
work, the main range of value is the 40A DC range. The 400A
range is good for working on my car (detecting the starter
motor in the car drawing 150A peak current). The thing has
peak hold, so you can find the "worst" loading during
a session if you want. The Hall Probe has a tendency to
wander, so you have to zero it with the zero button on
the front.

http://www.drillspot.com/products/88645/extech_instruments_corp_380941_digital_mini-clamp_on_meter

It also has AC amps, and I use one of the AC ranges for
working on electric motors. Like the 10A on my 230V
central air conditioner compressor.

That particular meter, isn't sensitive enough to work with
small currents. But for the range a PC might draw,
you can do some service work with one. For example,
on my old AthlonXP system, when running Prime95, I
might see 16A on the +5V rail (because back then, the
processor used to run the switching converter off +5V.
Not all boards did that, but some brands preferred to
use +5V instead of +12V. This was before the ATX12V
cable became popular.

That one claims to have true RMS measurement capability
(probably up to around 50KHz or so), but when attempting
to measure the distorted waveform from a non-PFC power
supply, it does a poor job. I have to conclude it
really isn't True RMS. If you want to measure input power
to the PC with reasonable accuracy, a Kill-O-Watt meter
is good for that, and also, relatively cheap. That's one
tool I don't currently have in my collection - but if one
was sitting in the hardware store, I could not resist an
"impulse buy" if it was available.

Paul

 

[Niel Humphreys]

Can an overheating ATX PSU blow the mainboard if the PSU fails?

I may be mistaken but ISTR there's a possibility of a voltage surge when
the PSU goes.

If it's an Intel based system then it is unlikely but if it is an AMD based
system it could well take the board and processor out with it.

 

[Johny B Good]

If you plan to repair a lot of PCs, I recommend a couple things.

1) Everyone should have a multimeter, because you can
get a basic one for $20. And it allows checking
the voltages. So bang for buck, it's a good deal.

+1 but it can't indicate high ripple and switching noise of course.

Measuring current flow, requires a breakout
cable of some sort - you need to be "in series"
when using one of those, to measure current flow.
Measuring all the rails would take all morning.

2) A second item, more expensive, is a DC clamp-on ammeter.
I can wrap the jaws of that, around wires of the same color
on the main ATX power cable, and make current measurements.
And see whether the load is normal or not. Or whether
the load is really too much for the replacement supply.
There are many AC clamp-on ammeters, but DC capable ones
are fewer in number. The DC ones use a Hall probe for
the measurement.

I got one like this, an Extech, before they got the ugly
color scheme. Mine is grey in color, throughout. For PC
work, the main range of value is the 40A DC range. The 400A
range is good for working on my car (detecting the starter
motor in the car drawing 150A peak current). The thing has
peak hold, so you can find the "worst" loading during
a session if you want. The Hall Probe has a tendency to
wander, so you have to zero it with the zero button on
the front.

I acquired a similar unit with 1200 and 200 A DC ranges. Despite the
high ranges, its saving grace for this sort of work was the fact that
it had a 4 digit display so could resolve to 0.1A.

The need for a zeroing button is an inherent feature of these DC
clamp meters. I just had a swift look at mine and it was showing
-127.4 A reading until I zeroed it. Even then I had to re-zero it
another 3 or 4 times before it stabilised.

This is something to watch out for when taking readings since you
have to double check the zeroing every few minutes to be sure your
reading is still valid if you need the accuracy to ascertain the total
DC power consumption to get a reasonably good idea of the PSU's
efficiency (when you've got a trustworthy AC watt meter to check the
mains input power).

http://www.drillspot.com/products/88645/extech_instruments_corp_380941_digital_mini-clamp_on_meter

It also has AC amps, and I use one of the AC ranges for
working on electric motors. Like the 10A on my 230V
central air conditioner compressor.

That particular meter, isn't sensitive enough to work with
small currents. But for the range a PC might draw,
you can do some service work with one. For example,
on my old AthlonXP system, when running Prime95, I
might see 16A on the +5V rail (because back then, the
processor used to run the switching converter off +5V.
Not all boards did that, but some brands preferred to
use +5V instead of +12V. This was before the ATX12V
cable became popular.

That one claims to have true RMS measurement capability
(probably up to around 50KHz or so), but when attempting
to measure the distorted waveform from a non-PFC power
supply, it does a poor job. I have to conclude it
really isn't True RMS. If you want to measure input power
to the PC with reasonable accuracy, a Kill-O-Watt meter
is good for that, and also, relatively cheap. That's one
tool I don't currently have in my collection - but if one
was sitting in the hardware store, I could not resist an
"impulse buy" if it was available.

Yep, the good ol' trustworthy Kill-A-Watt can be really useful to
check for unusually high (fault) loadings. I've got a couple of the
European versions and I'm quite impressed at their accuracy (compared
to some very early "Consumption Meters" I had the misfortune to buy
about 8 or 9 years ago). I have a MetraWatt analogue watt meter which
I bought a couple of decades ago which I use as the benchmark of
accuracy by which to assess these 'New Fangled' digital watt meters.

I'd certainly recommend you buy yourself an AC watt meter since this
can often provide all the information you need to assess what's going
on with a non-booting / no video display system. It's also handy to
avoid MoBo burnout when testing a box full of 2nd user CPUs for
shorted chips.

 

[Paul]

If it's an Intel based system then it is unlikely but if it is an AMD based
system it could well take the board and processor out with it.

I have to ask, what makes one more "prone" than the other ?

I haven't any first-hand experience with the problematic
Bestec, but I understand that one, damages a computer
because the +5V output goes to around +8V and damages
all the hardware in the system that uses +5V.

In the case of a "voltage surge", as proposed, the processor
has a switching converter (VCore) located around the
CPU socket. The switching converter provides conversion
and a degree of isolation, between +12V coming into it,
and the approximately +1V feeding the core of the processor.
And that is proposed as a protection mechanism for the
processor - the existence of another power conversion
stage. A few of the motherboard subsystems are like that,
having onboard regulation which can provide some
protection. The motherboard could still blow out,
if a voltage shoots high enough, but it doesn't
really matter what brand of CPU when that happens.
You will likely get to reuse your CPU and RAM,
as they're not directly exposed.

It would be nice to know a proposed mechanism for
this difference between Intel and AMD systems.

Paul

 

[Niel Humphreys]

I have to ask, what makes one more "prone" than the other ?


It would be nice to know a proposed mechanism for
this difference between Intel and AMD systems.

Intel CPUs have built in heat/voltage protection, AMD processors do not
which is why AMDs tend to be cheaper.

For instance, when the cooling protection fails on an Intel CPU it just
slows down and grinds to a halt until it cools down then works fine again.
With AMD they just fry and usually take the motherboard with them. The same
for voltage spikes etc.

 

[Jaimie Vandenbergh]

Er... no, not really for that reason. The economics of chip production
are complicated.

That might have been true about 10 years ago but not now. All AMD CPUs
since the Athlon64 have on-die thermal protection.

I'm pretty sure that AMD got it first, too - out of necessity, since
the early AMD's ran rather hot!

Cheers - Jaimie

 

[Paul]

That might have been true about 10 years ago but not now. All AMD CPUs
since the Athlon64 have on-die thermal protection.

My AthlonXP motherboard, it monitored diode temperature on
the processor. The processor has a thermal diode on
the silicon die (cathode and anode pinned out). The
motherboard designer connected that to an 8 pin chip,
and the output of the chip fed into the PS_ON# logic.

The end result was, if the heatsink fell off the processor,
my power supply would be shut off instantly. Since diode
monitoring is right on the die, there is no lag compared
to waiting for a socket thermistor to heat up. The latency
is zero. The eight pin chip requires no software,
so even if the BIOS crashed, the motherboard
still shut off. And that design, is an example of
correcting for a couple of faults with previous
protection concepts. No software needed to react.

It's true, a previous generation of S462 Athlon
didn't have effective protection. It might even
have been the case, that the thermal diode was
on the processor all along, but wasn't specified
in the design examples, as being necessary.
(The diode may have existed, for AMD staff in the
lab, to qualify cooling solutions.)

I think the later generations (S754, S939, AM2 and
so on), they have THERMTRIP. Which replaces the
eight pin chip my board had, and does it all inside
the processor. So now, a signal comes from the
processor, specifically for turning off the ATX
supply. The heatsink can fall off a S939,
and the power supply gets shut off when the CPU
is sufficiently warm.

The AMD website is so pathetic now, I no longer
try to find what pass for datasheets, to fill
in details. I can only assume that THERMTRIP has
been a standard feature to this day.

Throttling is another question. AMD supports P-State
changes, and software could change states on a mild
overheat. You don't want to rely on THERMTRIP,
because it's going to leave a dirty shutdown on
filesystems and need CHKDSK.

*******

A voltage spike is a completely different animal.

Since there can be switching converters for Vcore
and various DIMM voltages, those converters provide
a measure of isolation between the ATX PSU and
some of the plugin components. The motherboard
itself could be an easy victim, if any
motherboard logic runs off a rail directly.
So I cannot discount a Bestec-like event
taking out a motherboard, but your CPU and
RAM might be reusable.

The hard drive, because it connects directly to
the ATX supply, could also be a victim. The logic
board on there, has a protection device that
clips overvoltage. Like, if the +12V feeding
the motor rises to around +15V, there is a
protection device near the power supply terminals
that gets burned. That device is there to prevent
inductive spiking when a hard drive is
removed "hot". It is not intended to
crowbar an ATX supply, and cause the
ATX supply fuse to blow. So the hard
drive, one of the things that happens, is
the two protection devices (one on +5, one on +12)
fry, and the logic board, motor controller fry next.
Any sustained overvoltage, for periods of
seconds, will ruin the drive. The drive is
protected to some extent, against
millisecond long inductive spikes.
A Bestec, would just burn the protection
device on +5V.

A poster here, pointed out these things to me.
He posted that he had a hard drive, and a couple
components near the power supply connector were
burned. And using numbers off a similar logic
board, we figured out those components were
for overvoltage protection. So this isn't
something I read on a web site - it was
discovered while mulling over some visual
symptoms on a hard drive. (The only other
thing that likes to burn on hard drives,
is particular models of motor controllers )

Paul

 

[Jason]

Apart from the well known infamous Bestec ATX psus used by eMachines
models harking back a decade ago, most ATX PSUs fail without damaging
the MoBo.

The usual failure mode when SMPSUs overheat is for the switching
transistors in the HT module to go short circuit and smash the safety
fuse to smithereens in a flash of light accompanied by a loud pop and
a trace of smoke.

When a PSU shuffles off its mortal coil with such drama, only a
single modest transient voltage spike appears on the voltage rails
(the very small ferrite transformer can only transfer a limited amount
of power due to such a _single_ terminal pulse of current) which is
incapable of causing damage. IOW, the kindest failure mode wrt the
load is that dramatic sudden fuse shattering failure of the switching
transistors in the HT module.

There are other types of failure involving sustained overvoltage but
these tend to be rare unless the manufacturer goes out of his way to
bypass/ignore the overload/overvoltage protection features built into
all the standard smpsu controller chips that made it possible to
manufacture ATX psus at all in large volumes at affordable prices.

If the ATX PSU in question went "BANG!", then it's most likely that
no harm has been done and a drop in replacement ATX PSU will get
things working again.

I'm the OP for this thread. The fan on my ATX PSU seems to slow down at
times and makes a noise so I guess it's approaching the end of its life.
My PSU has not overheated yet and I'm still using the PC until I manage
to change the PSU or maybe just the fan.

Can you help me with this? ... If my PSU does overheat and might fry the
mainboard then would a surge protector on the incoming mains prevent the
PSU suddenly drawing mains power to feed the surge?

(I'm in the UK where I understand the mains voltage tends to be
reasonably stable.)

 

[>]

Can an overheating ATX PSU blow the mainboard if the PSU fails?

I may be mistaken but ISTR there's a possibility of a voltage surge when
the PSU goes.

Definitely yes, it can happen.

Bestec PSU's were renowned for destroying mainboards when they failed.

I've also seen PSU failures kill virtually every other part in the system.

I recall one system where only a PCI 56k modem and a floppy disc drive
survived.
HDD, DVD drive, CPU, memory, gfx card were all dead - even the cpu cooler
fan had died.

This is precisely why cheap n' cheerful PSU's should be avoided.

 

[Mike Tomlinson]

I'm the OP for this thread. The fan on my ATX PSU seems to slow down at
times and makes a noise so I guess it's approaching the end of its life.
My PSU has not overheated yet

How do you know?

and I'm still using the PC until I manage
to change the PSU or maybe just the fan.

It'll be too late by then. The PSU will have heated to the point that
the caps inside are damaged and no longer adequately smooth the output
voltages. When that happens, the motherboard caps, placed under
abnormal stress also begin to fail.

The time to change out your PSU is now, before motherboard damage
occurs.

would a surge protector on the incoming mains prevent the
PSU suddenly drawing mains power to feed the surge?

No, not at all. It deals with a surge on the *input*, not on the
output.

(I'm in the UK where I understand the mains voltage tends to be
reasonably stable.)

For now, but the Grid is approaching a point where backup capacity to
deal with unexpected outages (like Dungeness nuke station tripping
offline during last week's storms) is becoming marginal because so many
power stations, mainly coal burners, have been forced to close to comply
with Euro environmental bollocks.

This winter, expect brownouts at best and blackouts at worst. If no
more generating capacity is brought online next year and we can't import
enough power from the cloggies and frenchies over the interconnects,
blackouts *will* occur next winter.

 

[Mike Tomlinson]

Nope, Pentiums had it first. Tom's Hardware circa 2001 demonstrated
Athlons burning up and Pentiums surviving when the hs/fan was removed.

http://jasper.org.uk/THG_CPU_Cooling.avi

Amazing to think that was nearly 14 years ago.

 

[Johny B Good]

I'm the OP for this thread. The fan on my ATX PSU seems to slow down at
times and makes a noise so I guess it's approaching the end of its life.
My PSU has not overheated yet and I'm still using the PC until I manage
to change the PSU or maybe just the fan.

The fan is simply showing a lack of lubrication, try cleaning all the
dust and fluff (vacuum and a 1/2 inch paintbrush makes an effect
combination) before putting a drop or two of oil into the fan spindle
bearing (rubber plug hidden beneath the manufaturer's sticker).

This should, at the very least, extend the life of the PSU (assuming
the fan responds favourably - as they typically do - to such
treatment). You can check the PCB for any obviously 'blown' caps
whilst you've got the cover off although, ime, replacing any such caps
is far from guaranteed to extend the life of the PSU (a far cry from
the effect of replacing such caps on a MoBo where it usually extends
the service life by a year or three)

Can you help me with this? ... If my PSU does overheat and might fry the
mainboard then would a surge protector on the incoming mains prevent the
PSU suddenly drawing mains power to feed the surge?

If the fan stops completely, the most likely failure will be the HT
switching transistors going short (a flash and a bang) which is
generally not likely to trouble any of its rails with a voltage spike
worthy of such a description (more likely a voltage blip).

As for the use of a surge protector this won't make any difference as
far as your theory of failure mode is concerned. However, it can
reduce the risk of a voltage spike triggering an 'avalanche' failure
due to excess temperature reducing the collector/drain breakdown
voltage of the HT switching devices. Do you have a particularly bad
mains supply with regard to spike voltages?

(I'm in the UK where I understand the mains voltage tends to be
reasonably stable.)

You've got that right! Here in the UK you just don't get the same UPS
ROI benefit as they do in most parts of the US of A.

 

[Johny B Good]

I have to ask, what makes one more "prone" than the other ?

I haven't any first-hand experience with the problematic
Bestec, but I understand that one, damages a computer
because the +5V output goes to around +8V and damages
all the hardware in the system that uses +5V.

Wrong supply rail. It was the 5VSB rail that was susceptible to
overvolting. Typically the voltage would (presumably) creep up little
by little, reaching voltages in excess of 6v, typically 7[1] to 9
volts being observed ime.

Quite obviously, due to such overvolting failures only being observed
with these particular Bestec supplies, there was no effective
overvolting protection circuit on the 5VSB rail. An unfortunate
oversight since this powers all the standby circuitry of the MoBo (RTC
and PS Button monitoring circuitry) which remains exposed to the
steadily rising voltage even when the PC is shutdown whilst it
continues to recieve mains power (few users will go to the trouble of
switching off at the mains post shutdown).

In the case of a "voltage surge", as proposed, the processor
has a switching converter (VCore) located around the
CPU socket. The switching converter provides conversion
and a degree of isolation, between +12V coming into it,
and the approximately +1V feeding the core of the processor.
And that is proposed as a protection mechanism for the
processor - the existence of another power conversion
stage. A few of the motherboard subsystems are like that,
having onboard regulation which can provide some
protection. The motherboard could still blow out,
if a voltage shoots high enough, but it doesn't
really matter what brand of CPU when that happens.
You will likely get to reuse your CPU and RAM,
as they're not directly exposed.

That's a valid mechanism of protection for those particular items.

It would be nice to know a proposed mechanism for
this difference between Intel and AMD systems.

The AMD cpus up to and including the Socket A XP series had no built
in thermal throttling which I believe was introduced by Intel with its
P2/P3 processors (possibly only the socket 370 chips).

I don't think I ever came across any dead or shorted Intel CPUs ever
but I certainly saw a few dead or shorted AMD ones (outside of the box
of 6 or 7 2nd user socket A chips bought from a flea market trader for
a fiver out of which only two actually worked - most of the rest, as
expected, proving to be short circuit[2])

[1] 7 volts is the maximum Vcc (10 seconds) electrical stress limit
for TTL (and the CMOS drop in equivilents).

[2] Anticipating such MoBo destroying faults, I powered up with my
eyes on the Metrawatt Watt meter, switched to the 200v 1A range, and
my other hand poised to operate the PSU mains isolator switch in the
event that the wattage exceeded 120 (no disk drives, just the bare
minimum hardware required to allow a successful POST sequence to be
observed).

Testing the shorted CPUs produced a wattage reading that immediately
went north of the 200 watt mark on the scale (in fact pegging the
needle hard against the end stop) for the half second it took to
register the reading _and_ hit the off switch. Unfortunately, this
test carries too high a risk when using a digital watt meter, you
really do need an analogue meter for this sort of test.

 

[Paul]

I have to ask, what makes one more "prone" than the other ?

I haven't any first-hand experience with the problematic
Bestec, but I understand that one, damages a computer
because the +5V output goes to around +8V and damages
all the hardware in the system that uses +5V.

Wrong supply rail. It was the 5VSB rail that was susceptible to
overvolting. Typically the voltage would (presumably) creep up little
by little, reaching voltages in excess of 6v, typically 7[1] to 9
volts being observed ime.

Quite obviously, due to such overvolting failures only being observed
with these particular Bestec supplies, there was no effective
overvolting protection circuit on the 5VSB rail. An unfortunate
oversight since this powers all the standby circuitry of the MoBo (RTC
and PS Button monitoring circuitry) which remains exposed to the
steadily rising voltage even when the PC is shutdown whilst it
continues to recieve mains power (few users will go to the trouble of
switching off at the mains post shutdown).

In the case of a "voltage surge", as proposed, the processor
has a switching converter (VCore) located around the
CPU socket. The switching converter provides conversion
and a degree of isolation, between +12V coming into it,
and the approximately +1V feeding the core of the processor.
And that is proposed as a protection mechanism for the
processor - the existence of another power conversion
stage. A few of the motherboard subsystems are like that,
having onboard regulation which can provide some
protection. The motherboard could still blow out,
if a voltage shoots high enough, but it doesn't
really matter what brand of CPU when that happens.
You will likely get to reuse your CPU and RAM,
as they're not directly exposed.

That's a valid mechanism of protection for those particular items.

It would be nice to know a proposed mechanism for
this difference between Intel and AMD systems.

The AMD cpus up to and including the Socket A XP series had no built
in thermal throttling which I believe was introduced by Intel with its
P2/P3 processors (possibly only the socket 370 chips).

I don't think I ever came across any dead or shorted Intel CPUs ever
but I certainly saw a few dead or shorted AMD ones (outside of the box
of 6 or 7 2nd user socket A chips bought from a flea market trader for
a fiver out of which only two actually worked - most of the rest, as
expected, proving to be short circuit[2])

[1] 7 volts is the maximum Vcc (10 seconds) electrical stress limit
for TTL (and the CMOS drop in equivilents).

[2] Anticipating such MoBo destroying faults, I powered up with my
eyes on the Metrawatt Watt meter, switched to the 200v 1A range, and
my other hand poised to operate the PSU mains isolator switch in the
event that the wattage exceeded 120 (no disk drives, just the bare
minimum hardware required to allow a successful POST sequence to be
observed).

Testing the shorted CPUs produced a wattage reading that immediately
went north of the 200 watt mark on the scale (in fact pegging the
needle hard against the end stop) for the half second it took to
register the reading _and_ hit the off switch. Unfortunately, this
test carries too high a risk when using a digital watt meter, you
really do need an analogue meter for this sort of test.

I just found an analysis on this site. Amazing.
I didn't know any details from the Bestec era, just
that you should swap out a ATX-250-12E if you happened
to own one. They do a bit of a workup, here.

http://www.badcaps.net/forum/showthread.php?p=69599

The original and modified (hand drawn) schematics for
the +5VSB section, are copied here. The poster in that
thread, repairs the Bestec with a new design for the
+5VSB section.

"Bestec ATX 250 12E original.gif"
http://imageshack.us/a/img208/3773/9b7a.gif

"Bestec ATX 250 12E modified.gif"
http://imageshack.us/a/img96/2793/n8qj.gif

*******

"Death of a Gutless Wonder II: The Best of the Bestec"
http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story4&reid=154

"The 5VSB circuit proved to house a simple double transistor
switcher working in conjunction with... surprise... a bad capacitor
and no overvoltage protection circuitry."

The TPS3510 only monitors three rails, so can't be
relied on if there is a problem with the others.
The diode connection to Vsb, isn't for monitoring, and
is to keep the chip powered while the power supply
is in soft-off.

http://www.ti.com/lit/ds/symlink/tps3510.pdf

*******

For comparison, if we look at this schematic

http://www.pavouk.org/hw/en_atxps.html

the lower left corner shows a +5VSB implementation. What
I like about it, is it is followed by a 78L05, a linear
regulator. That might have protected the +5VSB rail a bit
better, but such a design would be underpowered for
modern systems (which run +5VSB at up to 3 amps). A
linear is highly inefficient, but that last chip
offers some protection against a few faults. It's
not going to be as good as a full featured +5VSB
designed by someone who cares.

*******

What the above doesn't square with, was at the time,
I was reading about hard drives being ruined by the
Bestec. If it was +5VSB that was failing, that
doesn't power the hard drive. The damage path
would be a lot longer, to get from +5VSB into the 5V rail
powered stuff. Back when I was reading up on this
stuff, there were reports of "5V going to 8V" and
"motherboard and hard drive ruined".

To explain what I mean by a damage path, I can give
an example from a lab accident. I was using a logic
analyser, on a system with an open faced power supply.
One of the test leads, fell into the supply (120V).
I heard a pop. I thought the analyser was destroyed
for sure. When the instrument was sent to our inhouse
service shop, the guy in there (a rocket scientist),
fixed it in two days, and only two ECL chips were ruined.
So unlike the Byte Magazine example where "all the tops
of the chips were blown off", contact with AC didn't
actually work its way too far into the instrument.
Some chips die, in order that others can live
Figuratively speaking... The more items a fault
has to go through, the more the effects get
attenuated.

Paul

 

[GS]

I'm the OP for this thread. The fan on my ATX PSU seems to slow down at
times and makes a noise so I guess it's approaching the end of its life.

If it's just the fan, you can replace that, Maplins sell case fans for
about a fiver IIRC, you just need to get the right size. I replaced mine
after it started making a continual buzzing sound. The only touble was the
original had been hardwired rather than connected with a plug but I was
able to connect the new one up without much difficulty anyway.

 

[Jason]

The fan is simply showing a lack of lubrication, try cleaning all the
dust and fluff (vacuum and a 1/2 inch paintbrush makes an effect
combination) before putting a drop or two of oil into the fan spindle
bearing (rubber plug hidden beneath the manufaturer's sticker).

This should, at the very least, extend the life of the PSU (assuming
the fan responds favourably - as they typically do - to such
treatment). You can check the PCB for any obviously 'blown' caps
whilst you've got the cover off although, ime, replacing any such caps
is far from guaranteed to extend the life of the PSU (a far cry from
the effect of replacing such caps on a MoBo where it usually extends
the service life by a year or three)


If the fan stops completely, the most likely failure will be the HT
switching transistors going short (a flash and a bang) which is
generally not likely to trouble any of its rails with a voltage spike
worthy of such a description (more likely a voltage blip).

As for the use of a surge protector this won't make any difference as
far as your theory of failure mode is concerned. However, it can
reduce the risk of a voltage spike triggering an 'avalanche' failure
due to excess temperature reducing the collector/drain breakdown
voltage of the HT switching devices. Do you have a particularly bad
mains supply with regard to spike voltages?


You've got that right! Here in the UK you just don't get the same UPS
ROI benefit as they do in most parts of the US of A.

Thanks for the info. I've cleaned the dust off (it was pretty clean
anway as I keep an eye on dust) and it didn't make any difference.

I hadn't realised I could lube the fan so I'll try that. While the PSU
is open I'll note the fan's fittings and get a spare one in case it's
too far gone for the lube to help.

Maybe I should also get a spare fuse. If the fuse markings are
indistinct, what's the typical fuse value for a 250W ATX PSU? (The
actual PSU model is FSP250-60GTV.)