Do I need a new power supply?

[SteveH]

<usual westom drivel snipped>

Number

one problem - a tech who forgot why those lessons in junior high
science apply - a tech that practices junk science. Reseating
connectors only cures symptoms; does not solve the problem.

Ah, Mr W.Tom, still talking bollocks I see.

SteveH

 

[Mike Tomlinson]

It cost the company maybe $200,000

[snip reams and reams of repetitive waffle]

Because they kept solving the problem by reseating connectors, then
they created a $200,000 loss.

Note how in w_twat's wacky world, that which starts out as "maybe $200k"
becomes "definitely $200k" many, many tedious words later.

No justification or accounting is presented for that figure. w_twat
made it up and later presents it as fact. This is w_twat's junk
science.

Hey Tom! When are you going to apologise for twisting my words where you
alleged that I told jasee to buy a replacement UPS when I had done no
such thing? You ****ed off out of that thread pretty smartish once you
realised your mistake, didn't you?

Are you going to answer these questions or not?

from "Bud~":

Still never answered - embarrassing questions:

- Why do the only 2 examples of protection in the IEEE guide use plug-in
suppressors?

- Why does the NIST guide says plug-in suppressors are "the easiest
solution"?

- Why does the IEEE guide say in one example "the only effective way of
protecting the equipment is to use a multiport protector"?

- In the IEEE example how would a service panel suppressor provide any
protection?

- Why does SquareD say "electronic equipment may need additional
protection by installing plug-in [suppressors] at the point of use."

For real science read the IEEE and NIST guides. Both say plug-in
suppressors are effective.

from me:

"You've claimed this on many occasions over many years [to work for a
"PSU shop"] but have never posted any evidence when challenged, for
example your claim that you have a degree. When challenged to provide
evidence of same (by anybody), you've invariably gone quiet or reverted
to personal attacks."

And this one (odd how my prediction came true, isn't it?)

"Explain how [to measure ripple voltage with a multimeter] in words of
one syllable, with numbers. (You won't, of course; you'll run away like
you usually do when challenged.)"

And this one?

"And how exactly do you put the PSU of a non-booting PC under maximum
load?"

And a new one?

"Can we please have a formal definition of a "computer grade" UPS?"

<fx: tumbleweed>

 

[Charlie]

It cost the company maybe $200,000 because the first and every tech
afterwards used a classic myth - reseating connectors - top ignore the
problem. Ignored the problem so long that the junk science created a
$200,000 expense.

I don't believe that I or anyone else in this thread was suggesting that
the original poster ignore the problem. Certainly, if reseating didn't
work (including "only worked for a short time") then it's time for a
different approach.

Those techs did what many ill trained techs so often do. They
changed things until something worked. They did not find the problem
before fixing it. And they assumed connectors can be fixed by
cleaning or reseating - which only happens when the defect is
elsewhere. IOW they 'fixed symptoms' rather than 'solve the problem'.

Reseating in the connector sometimes 'solves the problem', sometimes
does not and sometimes helps to narrow down the search to what is
causing the problem. In any case it is an easy and cost effective thing
to do.

You don't grasp the concept.

Not your concept.

Fixing things comes directly from
concepts taught in junior high science.

While I agree that people who understand the scientific method are
likely to be good at troubleshooting (because the thought processes
involve a similar logic), I believe this is getting too far off topic.

To have a solution means both
experimental evidence AND fundamental underlying theory. Without
both, that is classic junk science.

Oh well, if you insist on bringing science into a discussion that
doesn't need it:

A major tenet of the scientific method is that, when you propose a
hypothesis, the evidence you use to support it should be reproducible.
In other words, others should be able to perform the same experiments
and come up with the same results.
As an example, if a person proposes:

"If reseating a card fixes it, then the card is 100% defective."

Now ask yourself. Is this proposal scientific? I stated and so have
others that they have fixed computer problems by reseating connectors.
We cannot reproduce the results. In fact we have been shown no evidence
to support the claim. Therefore the claim is not scientific.
Normally, I'd say, so what, but it was you who decided to bring science
into a discussion so you'll have to live with it.

In this case, experimental evidence is 'reseating eliminates a
failure'.
Agreed.

Specs state that anything that 'reseating' would do must
never happen - in decades.

In other words the claim (specs) was not reproducible (not scientific).

Therefore both requirements for 'solving'
anything did not exist. If reseating fixes something, then electrical
reasons for why must also be explained.

Only if the fix was temporary as it was in this case. In many cases the
fix is permanent. The fact that the techs you know are incompetent, is
irrelavant to the discussion.

Reseating a connector means a
gross design failure or a defect located elsewhere.

We have already shown this claim to be 'not reproducible' (see above).

< snipped much talk about untrained people >

Much of what you claimed about reseating remains unsubstantiated by you
or anyone else. My personal experience shows your claim to be false.
Your rebuttal seems to center on tech people being incompetent which has
nothing to do with your claim. I think you know that, but in case you
don't, I would like to mention that there are competent people in every
field and there are incompetent people in every field. This holds for
scientists, engineers, designers, A+ techs, ditch diggers, toilet
cleaners, and people who make claims in newsgroups

Charlie

 

[westom]

the original poster ignore the problem. Certainly, if reseating didn't
work (including "only worked for a short time") then it's time for a
different approach.

If reseating fixes anything, then nothing is learned or solved.
Worse, the problem is now made harder to find. If reseating does not
fix anything, the tech is no closer to a solution - only wasted time
and energy. But reseating is what the ill trained will recommend
simply because they don't understand anything else. Worse, they did
not learn basic connector concepts.

Good diagnostics means using basic scientific methods. Same thought
process also made obvious that Saddam did not have WMDs. But many who
ignored good scientific thinking also blindly and foolishly believed
those WMD lies. It's not off topic. #1 reason to fix things is to
learn - to practice healthy thought processes. Also why every
military academy graduates everyone with engineering training. They
need people who can think properly; can deal with reality; can solve
problems. That means the same scientific thought process.

So what does *reseating a connector* do? If you have sufficient
computer knowledge, then your every post answered that question with
numbers. But datasheet and application notes make it obvious. Any
change is trivial. So trivial that if anything changes, then problems
exist elsewhere. I keep posting this. Where do you say, in technical
terms, what reseating does? What gets changed?

"A major tenet of the scientific method": the experiment says
nothing if the hypothesis does not provide (is not based in)
underlying principles - also called the underlying theory. Does not
matter how many times a magician can make the girl disappear (the
experiment). It does not prove transporters can beam people away. If
your hypothesis is based in myths or classic junk science assumptions,
then the experiment (your experience) says nothing. Reality demands
a hypothesis based in underlying principles. Reseating connectors is
junk science; is not based in any underlying principles. Manufacturers
even say why in numeric specifications. 'Reseating' is only proven
by the same 'scientific method' that proved Saddam's WMDs.

I will not rewrite long application notes and manufacturer specs
from connector manufacturers that you must have read before posting.
Your posts demonstrate you never read them. Your replies even
demonstrate contempt for how equipment is designed; so that connectors
not cause failures.

We who did this stuff even long before PC existed could immediately
spot poor techs. They would fix things by reseating connectors and
cleaning them with an eraser. Observation identified which techs were
inventing solutions because they could not understand the actual
reason for failure. Their analysis was devoid of numbers. We often
see the same poorly trained trying to solve electrical problems with
more chassis fans. 'Reseating' solutions are a classic symptom of
insufficient technical training. Insufficient technical knowledge is
why so many computer techs *solve* problems by reseating. They are
using 'experience' not tempered by underlying knowledge - basic
principles. Experience without underlying principles - junk science.

Electronics are designed so that connector limitations cause no
problems. When reseating solved problems, we knew 1) the actual
problem is not identified or remains, 2) the tech needs reassignment,
retraining, or a new employer, and 3) a fool trusts experience that
was not tempered by learning underlying principles.

Too few Americans grasp these most basic diagnostics techniques.
Too many just blindly know - shotgun - never bothered to first learn.
Saddam's WMDs are a classic example. Just another reason why the
Silicon Valley needs so many immigrants. Too many domestic computer
techs automatically know reseating connectors is a repair technique -
when the spec numbers and underlying design principles say otherwise.

I will not rewrite those long application notes from connector
manufactures that a learned tech has read. If reseating a connector
cures the symptom, a minor change created by reseating has only made
the intermittent even harder to locate. But an untrained tech will
immediately assume he has *fixed* it. No wonder Consumer magazines
had so many trivial problems unsolved by so many computer repairmen.
Computer tech even replaced a perfectly good power supply because
shotgunning is how junk science reasoning works. Reseating connectors
- just another example of shotgunning based only in wild speculation.

We watched the ill trained even clean connectors with an eraser.
And so that company suffered $200,000 in losses because their techs
were so naive as to fix electronics by reseating connectors. The
problem was not solved until the engineer saw bad workmanship
practices (reseating connectors), and then identified the problem.
The problem that a trained tech was being paid to find; but have been
perverted by myths. Instead the techs were "reseating connectors".
So ill trained as to cost the company about $200K.

 

[TM]

I've been getting the dreaded blue screen of death periodically. I ran
ram checkers on my Corsair 1 GB dual channel memory and everything is
fine. The BSO messages say device error or device driver error.

I get the BSO even while running Windows 7 with similar messages.

My configuration is:

Asus LG 775 MB
Pentium D 2.8 GHZ
512 MB x2 Corsair dual channel RAM
128 MB AGP 8x Video
Onboard Sound
17 inch LCD Monitor
1 Floppy Drive
1 USB keyboard
1 USB hub from monitor
1 USB extension cable
1 USB wireless mouse
1 CD/DVD ROM drive (8x)
1 x 250 GB Maxtor Drive 7200 RPM
1 x 250 GB WD 7200 RPM drive
2 x 500 GB Hitachi drive (previously 1x 1000 GB Seagate Drive)
1 400 Watt PSP PSU: +3.3V 28 OA, +5V 30 OA, +12V1 18OA, +12V2=18 OA,
+5Vsb=2.5A, 12V=0.5 A
(+3.3V& 5v) = 180 W, +12V1 & 12V= 348 W MAX.

I'll bet it's memory. Those so-called "ram-checkers" aren't worth squat.

RMA the Corsair sticks (if you want). Get a new pair of 2X2GB,

 

[Frustrated]

I'll bet it's memory. Those so-called "ram-checkers" aren't worth squat.

RMA the Corsair sticks (if you want). Get a new pair of 2X2GB,

I'm thinking it is the memory too. I have an old 512 MB stick of ram, I'll
install that and see how that goes. However, it could also be my
motherboard: Asus P2B-VM.

 

[Frustrated]

Very good tip....I completely forgot about the implications of defrag and
chdsk on the file structure on a drive that may not be working well. I've
been a victim of this once before.

Your useful replies will be from the few who actually understand what that
BSOD is saying. Who
are sufficiently informed as to not blame bad household current - a
classic myth - for such problems.

Your replies will only be as useful as the information provided.
That means posting those BSODs as Paul has requested. Otherwise
expect more responses from the least technically informed such as
blaming household current.

Numbers didnt provide any useful information when I cross checked them via
google, that's why they were not posted. For those inclined to make guesses
here are some logs:

Event Type: Error
Event Source: System Error
Event Category: (102)
Event ID: 1003
Date: 2009-07-25
Time: 8:17:55 PM
User: N/A
Computer: Windows
Description:
Error code 1000008e, parameter1 c0000005, parameter2 8057ea36, parameter3
f3b23c40, parameter4 00000000.

Event Type: Error
Event Source: System Error
Event Category: (102)
Event ID: 1003
Date: 2009-07-07
Time: 9:45:17 PM
User: N/A
Computer: Windows
Description:
Error code 1000008e, parameter1 c0000005, parameter2 bf835e1a, parameter3
f34dfbfc, parameter4 00000000.

Error code 1000007f, parameter1 00000008, parameter2 f7ac7d70, parameter3
00000000, parameter4 00000000.

Error code 1000008e, parameter1 c0000005, parameter2 bf835e1a, parameter3
f3123bfc, parameter4 00000000.

 

[John Doe]

If you are going to quote the eternal idiot "westom", please do it
more carefully. Thanks.

 

[Paul]

Your useful replies will be from the few who actually understand what that
BSOD is saying. Who
are sufficiently informed as to not blame bad household current - a
classic myth - for such problems.

Your replies will only be as useful as the information provided.
That means posting those BSODs as Paul has requested. Otherwise
expect more responses from the least technically informed such as
blaming household current.

Numbers didnt provide any useful information when I cross checked them via
google, that's why they were not posted. For those inclined to make guesses
here are some logs:

Event Type: Error
Event Source: System Error
Event Category: (102)
Event ID: 1003
Date: 2009-07-25
Time: 8:17:55 PM
User: N/A
Computer: Windows
Description:
Error code 1000008e, parameter1 c0000005, parameter2 8057ea36, parameter3
f3b23c40, parameter4 00000000.

Event Type: Error
Event Source: System Error
Event Category: (102)
Event ID: 1003
Date: 2009-07-07
Time: 9:45:17 PM
User: N/A
Computer: Windows
Description:
Error code 1000008e, parameter1 c0000005, parameter2 bf835e1a, parameter3
f34dfbfc, parameter4 00000000.

Error code 1000007f, parameter1 00000008, parameter2 f7ac7d70, parameter3
00000000, parameter4 00000000.

Error code 1000008e, parameter1 c0000005, parameter2 bf835e1a, parameter3
f3123bfc, parameter4 00000000.

http://aumha.org/a/stop.htm

1000008e, parameter1 c0000005

Bug Check 0x8E: KERNEL_MODE_EXCEPTION_NOT_HANDLED

http://msdn.microsoft.com/en-gb/library/ms794023.aspx

0xC0000005: STATUS_ACCESS_VIOLATION indicates that a
memory access violation occurred.

That means some code tried to access memory which was not set
up for that type of access. That is a safety feature. A possible
source of insanity is bad memory -- or malware.

1000007f, parameter1 00000008

Bug Check 0x7F: UNEXPECTED_KERNEL_MODE_TRAP

http://msdn.microsoft.com/en-gb/library/ms795478.aspx

0x00000008, or Double Fault, indicates that an exception occurs during a
call to the handler for a prior exception. Typically, the
two exceptions are handled serially. However, there are several
exceptions that cannot be handled serially, and in this situation
the processor signals a double fault. There are two common
causes of a double fault:

* A kernel stack overflow.
* A hardware problem.

*******

At the very least, I'd start with memtest86+, for at least a couple
full passes. Test 5 is where you may see errors reported.

http://www.memtest.org/

Keep logging errors, in the hopes that a driver name pops up.
If it isn't a memory problem, you're going to need all the
logging information you can gather.

One reason memtest86+ is good, is because of the percentage
of coverage. Memtest86+ cannot test BIOS reserved areas, so
there could be about 1MB worth of locations that cannot be
tested. Not everything the BIOS reserves, is used for something.
Other testing methods, give much poorer coverage. Running
Prime95 is a better test, from the degree of stress it
creates, but it lacks good coverage, and cannot test all
the memory. And in this situation, where the kernel is
reporting errors, Prime95 cannot grab memory that the
kernel is using. Memtest86+ can, because Windows is
not running when you're using a Memtest86+ floppy or
CD.

You can get Prime95 from here. p95v259.zip, Runs
a test thread per core. Tests 1600MB of my 2GB of
memory. You don't need to "join GIMPS" to use it.
I find this very good for general computer stability,
Vcore problems or marginal power supply. I may play
a 3D game at the same time, to add an additional test case.
I've had systems that are Prime95 stable, but adding the
game causes errors to show up and one of the Prime
threads gets an error.

http://www.mersenne.org/freesoft/

Paul

 

[John Doe]

You do know how to snip garbage, Paul?

 

[Frustrated]

Paul wrote:
::
:: http://aumha.org/a/stop.htm
::
:: 1000008e, parameter1 c0000005
::
:: Bug Check 0x8E: KERNEL_MODE_EXCEPTION_NOT_HANDLED
::
:: http://msdn.microsoft.com/en-gb/library/ms794023.aspx
::
:: 0xC0000005: STATUS_ACCESS_VIOLATION indicates that a
:: memory access violation occurred.
::
:: That means some code tried to access memory which was not set
:: up for that type of access. That is a safety feature. A possible
:: source of insanity is bad memory -- or malware.
::
:: 1000007f, parameter1 00000008
::
:: Bug Check 0x7F: UNEXPECTED_KERNEL_MODE_TRAP
::
:: http://msdn.microsoft.com/en-gb/library/ms795478.aspx
::
:: 0x00000008, or Double Fault, indicates that an exception occurs
:: during a call to the handler for a prior exception.
:: Typically, the two exceptions are handled serially.
:: However, there are several exceptions that cannot be
:: handled serially, and in this situation the processor
:: signals a double fault. There are two common causes
:: of a double fault:
::
:: * A kernel stack overflow.
:: * A hardware problem.
::
:: *******
::
:: At the very least, I'd start with memtest86+, for at least a couple
:: full passes. Test 5 is where you may see errors reported.
::
:: http://www.memtest.org/
::
:: Keep logging errors, in the hopes that a driver name pops up.
:: If it isn't a memory problem, you're going to need all the
:: logging information you can gather.
::
:: One reason memtest86+ is good, is because of the percentage
:: of coverage. Memtest86+ cannot test BIOS reserved areas, so
:: there could be about 1MB worth of locations that cannot be
:: tested. Not everything the BIOS reserves, is used for something.
:: Other testing methods, give much poorer coverage. Running
:: Prime95 is a better test, from the degree of stress it
:: creates, but it lacks good coverage, and cannot test all
:: the memory. And in this situation, where the kernel is
:: reporting errors, Prime95 cannot grab memory that the
:: kernel is using. Memtest86+ can, because Windows is
:: not running when you're using a Memtest86+ floppy or
:: CD.
::
:: You can get Prime95 from here. p95v259.zip, Runs
:: a test thread per core. Tests 1600MB of my 2GB of
:: memory. You don't need to "join GIMPS" to use it.
:: I find this very good for general computer stability,
:: Vcore problems or marginal power supply. I may play
:: a 3D game at the same time, to add an additional test case.
:: I've had systems that are Prime95 stable, but adding the
:: game causes errors to show up and one of the Prime
:: threads gets an error.
::
:: http://www.mersenne.org/freesoft/

Paul, thanks. I'll download memtest86+. I've used Prime95 before on this
machine and no anomolies were reported.

Any suggestions on a 3D game to run at the same time? Or perhaps run
bittorrent at the same time? I know bittorrent is really resource
intensive.

John Doe, is this better?

 

[westom]

Paul, thanks.  I'll download memtest86+.  I've used Prime95 before onthis
machine and no anomolies were reported.

One error code says a memory fault occurred probably in that used by
the OS. Failed memory address was bf835e1a. That may be a defective
location or where the error was detected when created by some other
defect. Second error implies the failure involves device drivers.

Your errors would come from a recently loaded and defective driver
(or other OS software), a bad memory location, power supply voltages
that are defective but can still run the machine, or other hardware
that is slowly failing, will only fail today when warmer, and will get
worse in the future.

AC power has no influence to the failure (which is why one wastes no
time posting insults).

You must confirm voltages are OK when system is under a maximum load
- from any one orange, yellow, red, and purple wire. Most important
would be the number from orange. Posting those numbers here may
provide further information you did not realize.

Memtst86 is another powerful. First execute in room temperature.
Then reexecuted at high temperatures that you do not like, but
computers love - such as a room at 100 degree F or by selective
heating with a hair dryer on highest heat settings. Heat is a
diagnostic tool. Selectively test by heating each memory board. And
by heating interface semiconductors such as the nearby Northbridge (a
large motherboard IC).

Common is for defective semiconductors to work at room temperature
but fail at 100 degrees. all seminconductors work perfecty happy at
temperatures you find uncomfortable.

Those three diagnostic information should result in useful
conclusions: heat, multimeter readings, and Memtst86.

 

[westom]

 My notes show me that over the years there have been numerous
times where reseating fixed the problem permanently.  I personally
believe that it should be one of the first things you should try when
you encounter flakey hardware problems.

We hit it with a hammer, and that also permenantly fixed it. We see
this often among those who know but did not first learn why.

Just because changing makes a temporary fix does not prove
anything. If the connector was not seated properly, then reseating
would fix it. But if the connection is fully seated, and reseating
fixed the connection, then the defect still exists - a symptom has
been cured.

Only because a fix was observed means nothing. We who fix things
the first time use well proven concepts to know why it fixes it.
Connectors are self cleaning when reseated. If minor electrical
changes in a connector causes failure, then the electronic circuit has
a defect. Instead the ill informed *know* the connector was fixed -
ignore the actual problem. Traditionally, that real defect only gets
worse with age. Reseating connectors is how to cure symptoms.

Those who know without learning why will know reseating connectors
work when curing symptoms is sufficient. If reseating the connector
elminates failure, a defect remains unsolved.

 

[John Doe]

We hit it with a hammer

Someone hit your head with a hammer, Tom.

 

[John Doe]

To the unaware reader...
Do not believe anything this idiot writes.

 

[Frustrated]

westom wrote:
:One error code says a memory fault occurred probably in that used by
: the OS. Failed memory address was bf835e1a. That may be a defective
: location or where the error was detected when created by some other
: defect. Second error implies the failure involves device drivers.
:
: Your errors would come from a recently loaded and defective driver
: (or other OS software), a bad memory location, power supply voltages
: that are defective but can still run the machine, or other hardware
: that is slowly failing, will only fail today when warmer, and will get
: worse in the future.
:
: AC power has no influence to the failure (which is why one wastes no
: time posting insults).
:
: You must confirm voltages are OK when system is under a maximum load
: - from any one orange, yellow, red, and purple wire. Most important
: would be the number from orange. Posting those numbers here may
: provide further information you did not realize.
:
: Memtst86 is another powerful. First execute in room temperature.
: Then reexecuted at high temperatures that you do not like, but
: computers love - such as a room at 100 degree F or by selective
: heating with a hair dryer on highest heat settings. Heat is a
: diagnostic tool. Selectively test by heating each memory board. And
: by heating interface semiconductors such as the nearby Northbridge (a
: large motherboard IC).
:
: Common is for defective semiconductors to work at room temperature
: but fail at 100 degrees. all seminconductors work perfecty happy at
: temperatures you find uncomfortable.
:
: Those three diagnostic information should result in useful
: conclusions: heat, multimeter readings, and Memtst86.

Considering I also get frequent BSODs with Windows 7 and have within a few
days of installing, would it be safe to say this is likely a hardware
problem?

How do I confirm the voltages are OK, especially from the orange wire?

 

[Frustrated]

John Doe wrote:
:: To the unaware reader...
:: Do not believe anything this idiot writes.

Why?

 

[westom]

Considering I also get frequent BSODs with Windows 7 and have  within afew
days of installing, would it be safe to say this is likely a hardware
problem?
How do I confirm the voltages are OK, especially from the orange wire?

Those BSOD errors limited the problem down to certain suspects (such
as Paul posted).

Using a 3.5 digit multimeter to measure that orange wire voltage (as
well as any one purple, red, and yellow wires). Simply set the meter
to 20 VDC. Touch wires. Read numbers. Then I have facts that result
in a useful reply.

Voltages are best measured when your machine is doing a much as
possible without crashing. For example, download from the internet,
while playing complex graphics (a movie), while searching the hard
drive, while playing a CD-Rom, etc. If unstable voltages are causing
your intermittent failure, then those numbers will expose that problem
- but only when computer is drawing a significant load.

As noted earlier, use Memtst86 also with semiconductors running hot
to humans (which is an ideal temperature to semiconductors). Memtst86
will not just report a failure. It will provide information to
isolate that problem to the memory or related motherboard hardware.
But first, something that define a failure must be found. Heat is a
diagnostic tool that often finds intermittents by temporarily making a
defect obvious.

 

[Frustrated]

westom wrote:
:
: Those BSOD errors limited the problem down to certain suspects (such
: as Paul posted).
:
: Using a 3.5 digit multimeter to measure that orange wire voltage (as
: well as any one purple, red, and yellow wires). Simply set the meter
: to 20 VDC. Touch wires. Read numbers. Then I have facts that result
: in a useful reply.

This may be a dumb question. But how could the voltages be measured just by
touching them when the wires are insulated?

 

[Paul]

westom wrote:
:
: Those BSOD errors limited the problem down to certain suspects (such
: as Paul posted).
:
: Using a 3.5 digit multimeter to measure that orange wire voltage (as
: well as any one purple, red, and yellow wires). Simply set the meter
: to 20 VDC. Touch wires. Read numbers. Then I have facts that result
: in a useful reply.

This may be a dumb question. But how could the voltages be measured just by
touching them when the wires are insulated?

To measure the voltage safely, do the following.

1) You need a multimeter, two test leads, and at least one alligator clip.
You place the alligator clip on the black lead, then snap the alligator
clip onto an I/O screw on the I/O panel on the back of the computer. This
ensures you don't get the tips of the probes too close together, and
short out the power supply. The chassis is grounded, and that is how
you'll be picking up ground.

2) Now, using the red test lead, with the meter set to 20V full scale or
a higher range, you push the test lead tip, into the back of the
main ATX power connector, where the wire goes into the plastic shell.
There is enough metal exposed in each nylon shell hole, for you to be
able to take a reading.

(slide the tip of the probe, next to where the wire goes in, in the right
most picture here. The wires are crimped to metal pins, and the metal will
be accessible on each one.)

http://www.playtool.com/pages/psuconnectors/main20pin.jpg

I don't know what you're trying to measure, but that is how to do it.
You can do that with the computer running. Since you've only got one
test lead in your hand, it is pretty difficult to get yourself into
trouble. Just make sure the red and black test leads are in the
"voltage holes" on the multimeter, and not the "current holes".

Paul