Prev: Can't erase DVD-RW - won't mount in drive once burned
Next: Economy build for home use.
From: Paul on 2 May 2010 19:50
daviddschool wrote:
> I had a power surge about 3 weeks ago - lost a drive, now making
> ticking noises. I had another two drive that seem to work
> sporadically now. So, I thought it might have been my power supply.
> Bought a tester and found out that all seems well. Next, I bought a
> new hard drive and found that it was also acting up - making noises
> like whirring up and down (like the other drives I thought were
> dead). So I am starting to wonder if it is my mobo - but how can I
> test this out?
>
> I am using a dual boot - Win XP and Win 7 - both drives work and don't
> work. Up and down. Right now my Win7 is working, but XP is not.
> Maybe both drives are on their last legs or maybe it is just the MOBO
> causing the issues - but now with a new drive, it is making the same
> noises, so that tells me that it is something other than the drives --
> I don't want to end up buying a new MOBO and find out it is something
> else again!
> Any suggestions?
> Also, all my info is backed, so I am ok with playing at this point,
> but I don't have another computer here to test this out on.

How about this for a test.

Disconnect the data cable, between the disk and the motherboard.
Do you hear "whirring up and down" now ? If you do, then
get out your multimeter, and check the 5V and 12V voltages.
Take a spare Molex disk drive connector, and probe there.

When I use a multimeter, I plug the black lead, into a screw hole
on one of the I/O connectors in the I/O plate area. The connector bodies
will all be at ground potential. Then, I only have to handle the red
lead, while making measurements. There is less chance of the meter
leads shorting together, if there is a large separation between
them. Plug your meter lead into the red pin and next on the yellow pin
on the spare Molex disk drive connector, and make your voltage measurements.

If the "whirring up and down" only happens when a data cable is
connected from the drive to the motherboard, that would be
more difficult to explain. If the motherboard were to request that
the drive be reset or restarted, that might not cause the motor
to spin down.

The disk drive controller board, monitors the voltage on +5V and +12V.
The controller only comes out of reset, when the voltages are stable.
I suspect, that conversely, if the voltages drop only a tiny bit
while the drive is running, you'll hear the drive reset itself again.
So while your power supply tester may have "passed" the ATX supply, you
should be double checking with a multimeter.

The spec for the ATX supply, is +/-5%. That means the 12V rail can be
between 11.4V and 12.6V. The supply will normally be more accurate
than that, and it shouldn't really be hitting those extremes. The
hard drive should not be resetting itself, with those kind of voltages.
But it would probably not take much more of a dip in voltage, before
the drive would reset, spin down and spin up again.

I've had the "whirring up and down" on a drive here. I had my AGP
video card and its Molex connector, on the same cable as a disk
drive. That drive started doing resets, as soon as I'd start a
3D game. By rewiring things, and using a separate cable for the
drive, the "whirring" stopped. So even excessive voltage drop in the
cable, can cause that. I suspect mine, may have been a low 5V rail,
rather than the 12V on the Molex being low.

Paul


From: daviddschool on 2 May 2010 20:18

> Disconnect the data cable, between the disk and the motherboard.
> Do you hear "whirring up and down" now ? If you do, then
> get out your multimeter, and check the 5V and 12V voltages.
> Take a spare Molex disk drive connector, and probe there.
>
> If the "whirring up and down" only happens when a data cable is
> connected from the drive to the motherboard, that would be
> more difficult to explain. If the motherboard were to request that
> the drive be reset or restarted, that might not cause the motor
> to spin down.

It is interesting that you mention this because when I disconnect the
data cable, I was still hearing the whirring up and down of the drive
and I thought it was odd. I don't have the know-how to use a
multimeter on my computer.
More interesting is that I am using WIN 7 on a 1TB drive right now on
the machine. This drive was unusable a little while ago, but seems to
be fine now. My XP drive is unusable now - not sure why. And the new
drive I bought (which I have taken out of the machine), is sitting
waiting - hopefully still in working order. I don't have a new PSU
now, so it is really hard to test the whole setup.
From: Paul on 2 May 2010 23:20
daviddschool wrote:
>> Disconnect the data cable, between the disk and the motherboard.
>> Do you hear "whirring up and down" now ? If you do, then
>> get out your multimeter, and check the 5V and 12V voltages.
>> Take a spare Molex disk drive connector, and probe there.
>>
>> If the "whirring up and down" only happens when a data cable is
>> connected from the drive to the motherboard, that would be
>> more difficult to explain. If the motherboard were to request that
>> the drive be reset or restarted, that might not cause the motor
>> to spin down.
>
> It is interesting that you mention this because when I disconnect the
> data cable, I was still hearing the whirring up and down of the drive
> and I thought it was odd. I don't have the know-how to use a
> multimeter on my computer.
> More interesting is that I am using WIN 7 on a 1TB drive right now on
> the machine. This drive was unusable a little while ago, but seems to
> be fine now. My XP drive is unusable now - not sure why. And the new
> drive I bought (which I have taken out of the machine), is sitting
> waiting - hopefully still in working order. I don't have a new PSU
> now, so it is really hard to test the whole setup.

OK, if you lack the skills to measure the power supply, do
you know how to change out the supply and test with another ?
Is the supply a standard ATX type ? The wire color code can
be found in one of the ATX specs. For example, in the third
spec, the colors for the main connector are on PDF page 37
(section 4.5.1).

http://web.archive.org/web/20030424061333/http://www.formfactors.org/developer/specs/atx/ATX_ATX12V_PS_1_1.pdf

http://www.formfactors.org/developer/specs/atx/ATX12V_1_3dg.pdf

http://www.formfactors.org/developer/specs/ATX12V_PSDG_2_2_public_br2.pdf

Those specs are chronological, with the specs being for:

1) 20 pin main connector, with -5V still present on the connector.
Maybe you'd find those about 10 years ago. In some cases, the
power supply has -5V on it, but the motherboard isn't actually
depending on that voltage.

2) 20 pin main connector, with -5V removed

3) 24 pin main connector, with -5V removed

So one of those specs, could be compared to your main connector.
(Naturally, the color of the wires is the same on each one, on
the corresponding pins. It is not like they changed the order
of the colors on each one or anything. Each spec is a slight
tweak from the previous one. The 24 pin spec, adds 4 pins, without
changing the colors on the existing 20 pins.)

There is a page in each spec, which lists the wire colors on
the main connector. The wire color implies a certain rail
voltage in each case. If the wires weren't in a standard
color configuration, you might suspect a non-standard supply.
There was an era, when that was popular. Later, the main
PC manufacturers came to their senses, and started using
the same supplies everyone else (like home builders) did.
But it remains as a task for people doing computer repairs,
to verify that they're not dealing with that problem.

To replace a supply, all that is needed, is a screwdriver
for the four screws holding the supply. And a careful
methodical approach to removing it. The first time I did
it, I made diagrams of how everything went back. If the supply
is a standard ATX, then all you have to ensure in that case, is
that you have enough cables on the new supply, that the wires
are long enough. For example, say your old supply had two
cable runs for disk drives, with three Molex on one, and two
Molex plus a floppy connector on the other. New supplies will
be peppered with SATA power connectors, which may not be useful
to you. So you want to check that either the new supply
has more than two cable runs, of which two cable runs are
useful configurations. Or, perhaps, you get a power supply
with modular cabling, and verify that there are some cables
present which have the hard drive and/or floppy power connectors
that you need.

You can use the model number of your pre-built computer, and
either use a general search engine, or use Ebay, to search
for a replacement supply. If there is a match, you can be
guaranteed that any prospective vendor, will not tell you
whether the supply is standard or not. If it was standard,
you'd simply run away and find a supply which is cheaper
than the one they're selling. So don't expect technical
advice, from someone claiming to sell a replacement
for your computer. Even the original computer manufacturer,
may not offer any details about power supply type. They
want you to buy their overpriced product.

If you've already replaced the motherboard in the computer,
with a retail (home builder) motherboard, then that tells you
a standard ATX supply must have been in the computer case
already. If you have the original motherboard in the pre-built
computer, the manufacturer could do whatever they wanted, and
then you have to be more observant, to check the type.

Some power supply testers, may use a "window comparator" on
each rail for testing. For example, two analog comparators,
one set to 11.4V and one set to 12.6V, could be used to test
the 12V rail. Then, a logic equation,

"is voltage > 11.4 and voltage < 12.6 ?"

is evaluated, to control a LED which says whether that rail
passes or fails.

If the power supply tester doesn't make allowances for the
motherboard to be connected at the same time, then you're not
doing a "load test". You can't tell whether a power supply is
weak, unless some kind of representative load is present. That
is why using a multimeter can be a useful exercise. For example,
by connecting the black lead to a rear I/O screw (some have a
hole in the screw body, which will provide a place for your
meter probe), then connecting the red probe to the main wiring
harness, you can check voltages while the computer is running.

The main power connector has a nylon shell for the pins, but
there is exposed metal in each "cell" defined by the nylon shell.
You can touch the tip of a meter probe, on the metal within each
cell, while the computer is running. That is how you can verify
the voltages with a load in place. The connector is not so
carefully insulated, that you cannot get a measurement. I'm
convinced someone did that on purpose, because it would have
been easy to "super-seal" the whole thing up, so that
no meter probe could get in there, if they wanted. They could
make a molded connector, with little extra effort or cost.
So the current design, does seem to support measurement
while the system is running. You probe, where the wires
go into the nylon shell. And by holding only one probe in
your hands, you can make measurements with no risk of
shorting to the other probe.

I have a load box I built for testing power supplies, which
is how I can verify a power supply before I use it. That
would cost you anywhere from $50 to $100 for some power
resistors. Mine draws somewhere around 100W, and is not
intended to "crush" the supply or ruin it. An idling
computer may draw that much power. I let the supply
drive that load for a couple hours, and take measurements
at the end of the test, to verify the supply is OK. That is
how I test a new supply, if I buy one, without connecting
it to the motherboard. I place a fan next to the power
resistors, to help keep them cool during the test. I don't
expect everyone to rush out and do that, but that is
another alternative arrangement for testing.

HTH,
Paul
From: daviddschool on 3 May 2010 20:09
I ordered the new PSU - should have it in 3-5 days, but now I am
getting memory dumps or something while Win7 is running. Really
weird, taking more time to load properly each time and the BSD
occasionally will haunt me but it goes so quick I can't read it. Any
ideas ? Does that still sounds like the PSU?


From: Paul on 3 May 2010 20:30
daviddschool wrote:
> I ordered the new PSU - should have it in 3-5 days, but now I am
> getting memory dumps or something while Win7 is running. Really
> weird, taking more time to load properly each time and the BSD
> occasionally will haunt me but it goes so quick I can't read it. Any
> ideas ? Does that still sounds like the PSU?
>

That is a separate and new symptom.

The state of the RAM, will not affect "whirring" on a disk drive.

To test memory, you can use memtest86+. It is available for
floppy, CDROM, and flash stick. They even include it on some
Linux LiveCDs, as a boot time option.

http://www.memtest.org/

For the CD version, you download the ISO, and use a program like
Nero to convert the ISO9660 file into a bootable CD.

http://www.memtest.org/download/4.00/memtest86+-4.00.iso.zip

The test will run forever, until you stop it. Pressing the
<esc> key, will trigger a reboot. For quick tests, I recommend
two complete passes. If you're in a rush, typically advancing
the test number to "test 5", may show you whether the memory sticks
are throwing errors. No errors are acceptable. Sometimes, if the
error count is low (1 or 2 errors), bumping up the VDimm voltage
one notch may be enough to fix it. Or relax Tras by increasing its
value upward one notch.

In terms of a power supply failure damaging stuff, some subsystems
are more exposed than others. A failed power supply shouldn't generally
damage RAM, because there is a regulator on the motherboard between
the power supply and the RAM. The RAM doesn't run directly from the
power supply. If the disturbance caused by a power supply problem is
moderate in nature, the regulator may feel it, but the RAM may be shielded
from it.

I've had several RAM failures here. In one case, I had a failure
roughly one month, after changing from two sticks to four sticks.
I have to conclude from that, that the RAM may have operated at
elevated temperature, due to poor cooling. So that to me, is a possible
factor. But I've had two batches of locally purchased generic RAM,
which failed just outside the warranty period, and that stuff was
not abused in any way. (it had good cooling and wide spacing between
sticks.)

It's your choice as to what testing you do, and when. Personally,
I'd hold off until the new power supply is in place. And then you
can go through your test cases again.

You can also run Prime95 as a test program. If memtest86+ is testing
clean, then I'd give Prime95 a shot. It runs within Windows or Linux,
as an application.

It runs a test thread per core, and tests both memory and processor.
It uses a set of math equations with a known answer, so the program
can detect rounding errors, or other kinds of errors. The test won't
tell you which bytes on the DIMM are bad, but if the test runs for
hours on end, you usually conclude from that, that the computer is
ready to use for serious work.

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

http://mersenneforum.org/gimps/p95v2511.zip

To use that program, you don't have to "join GIMPs". There is a prompt
that appears, when you start the program, but you don't have to bother
with that. You're "just stress testing". If you want to join the search
for large Prime numbers, that is fine, but isn't essential to using
the program for stress testing. On my machine, with 2GB of RAM,
the program will test around 1.6GB or so of RAM (it can't test the RAM
being used by the operating system). You can reduce the
memory setting, and leave enough memory for other programs to run.
For example, I sometimes play 3D games, and run Prime95 at the same
time, as a stress test.

Paul
First  |  Prev  |  Next  |  Last
Pages: 1 2 3
Prev: Can't erase DVD-RW - won't mount in drive once burned
Next: Economy build for home use.