micky said:
Triple failures? Hard Drive Data? Hard drive? Video Card? CPU?
Power Supply?
Last night, when I was running, as I have for at least 6 years, a
home-built computer with 1.5Gig memory and an 800MHaz CPU. All PATA,
no SATA.
I'm ib a another computer now and don't have details about the broken
one, but I can probably find them.
I clicked Back on my Firefox3 and the compuer restarted.
(It got to the point where it urges me to run chkdsk, but I was in
the middle of something and skipped that. The screen went black (or
blue) as normal, but before any Windows screen showed up, it restarted
again.
This time it offerred me safe mode etc. but I chose regular. I let it
run chkdsk and all it reported were two lengthy errors in avg log
files. But when it got to the same place, it restarted again.
This time I chose Safe Mode with Internet and it started fine, and I
even retrieved a couple of things from Usenet.
I wanted to go back to regular mode so I restarted.
This time it only got as far as the memory count-off, and it stopped.
Normally it shows a grid about my hard drives and CD dirves.
When I pushed the restart button, it took me to the CMOS/setup page.
Looking there, I found in the Boot Order, for hard drive it was set to
None, and had only two options, None and Skip. But the CD options
were both of my CD drives. A network was set for the fourth place to
look even though I don't remember ever doing that.
In the first page of setup, I have all four possible PATA drives set
to Auto, so iirc there woudl be no indication there about what drive
it finds.
I Escaped out, checking, Don't save any changes. Usually that is
plenty on other times, once or twice a year, when I get kicked to the
Setup screen.
It only got to the memory count-off again, Reset again put me on the
Setup screeen.
Turning it off completely and restarting showed a black screen. Even
though the red light of one CD drive goes on for a moment (I didn't
see the second one's light)
The screen is black partly because the monitor never wakes out of
standby, and if I turn it off and back on, it displays a little square
in the middle of the screen which says, Standby, Wake with PC, or
something like that.
It was very low humidity last night so I guess it was hotter than I
thought, mauybe 85, and for the first time I had five Firefox Windows
open, with at least 100 tabs. I had turned the fan/temp monitor off
for the winter.
I would think I overheated the CPU but it did start fine in Safe Mode,
and I sort of think some other chip is responsible for the POST
routine.
Waited 6 and 15 hours and works no better than it did.
I have I burned out my CPU, my video card**, the harddrive, my power
supply?
**The video card has acted up 5 times ln the last year, with a red
colored ghost for each letter and lots of funny dots on the screen to
where it's impossible to read. . If I put the computer in standby for
even a couple minutes, it's been all right again for another 6 hours.
I have a back up of the hard drive, but of course it's tailored for
this computer. I'd rather fix this computer then try to copy piece by
piece from the harddrive only.
I can fix this if it's not the CPU, but helpful advice would be much
appreciated.
Thank you.
[Based on the attempts in the header of one of the posts,
crossposted to alt.comp.hardware . I think that is where you
meant it to go.]
There are a few things you can start with.
1) Visual inspection. Use your senses, to detect something is amiss.
2) Simplify hardware setup and retest.
3) Use the PC "beeper" function for hardware testing, if
video no longer works well enough to get error messages
from the PC.
For visual inspection, you take the side off and look for loose wires.
Or, perhaps a strap on a heatsink broke, and the heatsink is hanging down.
Check, with power off, that fans turn freely. And that heatsinks are not
clogged with dust.
Now, to the hardware specifics. You want to inspect the "capacitors" on
the motherboard, for bulged tops or brown or orange colored deposits on
top, or near the base of the capacitor. The capacitors are aluminum cylinders
with a plastic sleeve, with the component value printed on it. On the top
of the cylinder, there is a "stamped" pattern, perhaps the letter "K",
where the seams in the metal are for pressure relief. If hydrogen gas
builds up inside the capacitor (it is failing), the seams open releasing
the pressure before there is an explosion.
Your computer predates the "capacitor plague" incident. So your capacitors
might not be failing prematurely. Capacitors may last for fifteen years or
more, if the internal chemistry is still good. The capacitor plague ones,
lack stability internally, and chemical breakdown can happen, cold, over
a period of a couple years (I've had some fail like that).
http://en.wikipedia.org/wiki/Capacitor_plague
The top of the caps should be flat, and the pressure relief seams should
still be intact. If the capacitor has failed, a "slight" failure causes
an increase in Vcore ripple. If the capacitor shorts out internally,
you may see smoke, hear "sizzling" at startup and so on. If a cap fails
near the processor socket, it can take out a toroidal coil of wire
or one or more MOSFETs (things with three legs soldered near the CPU
socket). While the Vcore regulator is protected against some fault
types, collateral damage can still occur.
Inside the power supply, you can have similar issues. If your power supply
was a replacement, purchased in the last five to seven years, it could be
it was affected by bad caps as well. I had an Antec power supply (contract
manufactured by ChannelWell CWT), that failed pretty much exactly
like this one. You can see orange deposits on at least four caps here.
http://upload.wikimedia.org/wikipedia/commons/8/84/PSU_Caps.jpg
It's only advisable to inspect the power supply, if the warranty
has expired. Unplug the supply. Remove four screws on the top. Remove the
top plate, noting how any insulation sheet fits into place. Put any
insulation back, exactly as you found it. You may "look but don't touch".
This is an inspection only. If you see the orange deposits, you'll need
a new supply (unless you live in a country with a ready supply of good
quality caps). The reason you "can't touch" in there, is the main cap
is charged to 300VDC or so, and if any juice remains on there, it's
dangerous. You always assume the bleeder resistor has failed, and the cap
is fully charged. Put the cover back on when you're finished. (The reason
you can't do this with a warranty in place, is usually there is a tear-away
sticker over one of the screws, so the manufacturer can detect tampering,
and refuse warranty service if the supply has been tampered with.)
So you're looking for visual clues, loose wires, broken heatsink or fan,
"bad caps", the smell of burning or chemical smells and so on.
The motherboard may have a coin cell to power the CMOS (BIOS settings).
If you use a multimeter, and ground the black lead to a screw on the
chassis (like a screw on an I/O connector), you can probe the top of the
CR2032 coin cell and check the voltage. It should read around 3V. Below 2.3
or 2.4V or so, the battery can fail to maintain BIOS settings when all
power is off on the PC. I didn't see anything in your symptom
description to implicate the battery.
Your startup where disks weren't detected, could have resulted in
the BIOS modifying the boot order. BIOS designs have evolved over
the years, in how they handle the boot order. Time was, the BIOS was
pretty good at keeping a reasonable boot order. I have some PCs here,
which are terrible at it, and the BIOS frequently does whatever the
hell it feels like. And puts the disks in some order I didn't specify.
This is normally triggered by a change in the hard drive configuration.
*******
In the "Simplify" department, you remove stuff according to the symptoms.
If you can no longer get video, you'll have to unplug a fair bit of stuff.
To offload the power supply, you can unplug the hard drives and CDROM drives.
Disconnect both the power cables (Molex 1x4) and data cables, making note
of the orientation and location of the cables for later. Leave the
floppy connected, so you can run memtest86+ later.
If you have no video, you can run "beep tests". These rely on a working
computer case speaker, connected to the SPKR 1x4 header pins on the
motherboard. If you were to unplug an AGP or PCI video card, such that
there was no video, the BIOS would "beep" a three beep pattern, indicating
a video failure. This has the beneficial side effect, of proving the
processor works, it read some BIOS code, it carried out tests. So in
fact, any beeps heard, are a positive sign.
If you power off again, and remove RAM sticks (placing them in an
antistatic bag), you can repeat the beep test again. A different pattern
will be heard. Video and RAM failures use two or three beep patterns,
and the pattern really isn't important, except that each kind of failure
uses a different pattern. Since the "success case" (normal PC startup)
is a single beep, any other beep pattern means the BIOS has discovered
a fault.
If you get zero beeps, then the processor might not be running BIOS
code. Occasionally, on an older computer, the BIOS chip gets "bit rot"
and the code is corrupted. But in your case, this didn't start with
a BIOS level failure - your computer was running at the time, so it
isn't likely to be a bit rot problem.
Your PC description says "800 MMz". That could be 800 MHz Pentium III
or Celeron. That would be powered by the main 20 pin power cable.
It wouldn't likely have the auxiliary power cable ATX12V with 2x2
power connector on the end. The other interpretation of your
description, is you have an Athlon processor. That could be
powered from the main cable as well. I have an AthlonXP motherboard,
and it doesn't use ATX12V, so up to 65W is drawn from the 5V rail of
the main power cable. (Athlon PCs may need a pretty decent +5V
current rating on the power supply. Not all modern power supplies
are a good match for that.)
The Pentium III and Athlon, may have BIOS monitoring of CPU temperature.
The BIOS is capable of a relatively slow response to overheating. If
a heatsink falls off an Athlon, it can overheat so quickly, the
processor crashes before the BIOS code can switch off the power.
Then, the power stays on and the processor is cooked. It was in
later generations, that "hardware overheat protection" in the
form of THERMTRIP was added to PCs, causing the power supply to
shut off, if a monitoring diode on the CPU silicon die is
overheating. But that happened a bit later perhaps, than your
800 MHz CPU.
In terms of probabilities, the power supply is a fairly unreliable
component. If you had no diagnostic equipment, and just played
the part of an equipment swapper, you'd swap that out first.
The CPU is relatively reliable.
RAM, isn't as reliable as they'd like you to believe. I had yet
another failure in that department, a week ago. (Fortunately,
not on one of my good PCs.) The computer vintage was around the
same age as your machine. I had (3) 512MB PC133 SDRAM in the
machine. Two sticks had failed. I had bought a grand total of
eight sticks of that stuff. What is funny, is the five remaining
sticks all tested good. Two of three sticks sitting in the PC
for years, were the ones that failed. It almost suggests some
kind of metallurgy problem while in storage, as that PC doesn't
get to run very often.
As part of your "beep testing", you remove video card and RAM,
and listen for beeps. If you hear two beeps, then maybe the processor
is still running. Then, you add in one stick of RAM (with the
power cord unplugged, no power present). Now, does the beep
pattern change to three beeps (video card missing) ? If so,
then the RAM might be OK. If you heard zero beeps, after
adding back the RAM, it could be that locations below 640K
have gone bad. You'd power down, and try one of the other
RAM sticks from your collection, trying them one at a time
until you find a (relatively) good one.
Say you install one RAM stick, hear the "missing video" beeps,
add in the video card, and now the machine POSTs again. You can
leave the hard drives and CDROM disconnected, and boot a memtest86+
floppy for a test. Now, you can verify memory locations above 640K.
http://www.memtest.org (scroll half way down, for downloads)
Do at least one full pass, with that test program. If you own
multiple RAM sticks, you can test each one individually, to
determine if any of the sticks are bad. Install any RAM that
completes a full pass. Once you've weeded out the duds, you
can do one final test run with all candidate sticks reinstalled.
Always fully power off, before making hardware config changes.
So now we've got RAM, video, floppy, but no hard drives. Now
you can cable those back up, and look for them in the BIOs
detection screen. Are they still missing ? Test your storage
devices one at a time (being careful to properly jumper the
drive for whatever cable config you're using to test). If
two drives disappear, and they share a common cable, you'll
need to test the drives one at a time, either as "Master" or
"Cable Select", depending on the ribbon cable type (80 wire
cable preferred).
Anyway, that's a few ideas to try out.
HTH,
Paul
