Numerous Hardware Issues (and perhaps some software issues, too)

[Daave]

I'll start at the end.

Last night I left my PC on. I was downloading the installation file for
Ultimate Boot CD for Windows. The download apparently had completed. And
then at some later time, the PC seemed to freeze. Since I had woken up
in the wee hours of the morning, I went to the PC (at about 4:15 AM) and
I noticed that the screen was frozen: not only could I not move the
cursor with my mouse, but the time display in the system tray was frozen
at 3:40 AM.

I assume something significant happened at 3:40 AM.

Everything *looked* okay. The lights and fans were working. The little
light next to where the LAN cable plugs in showed activity. Nothing was
hot. I noted the temperatures in Abit's uGuru utility, and they were
fine.

I decided to run the Windows Memory Diagnostic (just the Standard
tests). In the first pass, there were failures. I think this is
significant. (The significance will become more apparent when I discuss
the history below.) But I'm not sure if this points to memory errors or
possibly something else (e.g., something wrong with the motherboard).

FWIW, there was nothing significant in the Event Viewer.

I also ran the SeaTools hard drive diagnostic, which was negative.

Just for yucks, I disconnected the printer and external hard drive (both
USB devices). I rebooted and ran WMD again. After nearly 2 1/2 hours and
three passes in Extended mode, there were no errors!

So far, here are my ideas as to what is causing this issue:

1. Since this behavior has *not* happened with either USB device plugged
in (not yet, anyway), there might be an anomaly associated with one of
them (probably the external hard drive) that is responsible. I somehow
doubt this is the case, but I figured I'd throw it out there, anyway.

2. My wife's personal care attendant was moving stuff around the other
day, and I believe the PC (which is on the floor) was moved slightly
while it was running. Whatever happened as a result might be the cause
as these hardware problems occurred after this incident.

3. BIOS settings are not quite right or there is something wrong with
the CMOS chip or perhaps motherboard battery (which I haven't replaced
just yet). Here are the specs of my PC:

- Abit IX38 Quad GT motherboard

- Intel Core 2 Duo E8400 CPU

- Crucial Ballistix PC8500 DDR2 RAM (2 x 1GB)

- 250 GB Seagate ES.2 SATA2 hard drive (this is the one that has Windows
XP, SP2)

- 500 GB Seagate 7200.11 SATA2 hard drive (for storage)

- Lite-On 20x DL DVD +/- RW drive

- A-922 case

- 600 watt Tiger power supply

- MSI 512MB HD3870 OC PCI-E graphics card

Regarding this last item, when I originally built the system this past
April, the connection going into the card was not 100%. I still haven't
gotten a newer monitor; I am using my old Samsung CRT monitor which
requires an adapter. Once the connection was more solid, all was well
(for over 4 months). Just to be on the safe side, I checked this
connection again. I also checked that the card was seated properly.

Here are some of the BIOS settings:

Frequency: 3230 MHz
CPU Operating Speed: User Define
Multiplier Factor: 9.5
Estimated New CPU clock: 3230 MHz

This last item is in gray and has an X next to it. It seems as if this
is overclocked. If so, I wonder how this happened as I never defined
these settings. Might this be the issue?

Other gray lines with Xs:

Target CPU core voltage: 1.2250 V
DDR2 Voltage: Auto
CPU Vit Voltage: 1.10 V
MCH 1.25V Voltage: 1.25 V
ICH 1.05V Voltage: 1.05 V
ICHIO 1.5 Voltage: 1.50 V
DDR2 Reference Voltage: Default
CPU GTLREF 0 & 2
CPU GTLREF 1 & 3

Nothing seems out of the ordinary here.

Also, CPU fan speed (under Fan Speed Monitoring): 1680-1740 RPM.

In Integrated Peripherals, On Chip SATA device:

SATA mode: IDE (other choices are RAID and AHCI)

Speaking of SATA, when I installed XP, there was the option to press F6
if I needed to install SCSI or RAID drivers. I never did this because I
do not have a RAID setup. And the installation went fine. Should I have
installed SATA drivers from the get-go? Or did I do it correctly?

Also, in the options section of the BIOS is an option to load fail-safe
defaults. I haven't done this yet. Is it useful to try?

Oddly, another time I went into the uGuru utility, the values for the
CPU had changed! They were:

CPU Operating Speed: 3000 (333)
Multiplier Factor: 9
Estimated New CPU clock: 3060 MHz

So, I'm not really sure what's going on here.

Finally, since I mention ACPI below, the BIOS value for ACPI Suspend
Type is S3 (Suspend-To-RAM). The other choice is S1 (Power On - Suspend)

4. There is still a loose connection I have overlooked.

5. There might be something wrong with the motherboard.

6. There might be a problem with the power supply. Although I do have a
multimeter, I haven't used it yet. But that might be one of the next
steps.

7. A power surge. Although I have a power strip that is supposed to
protect against power surges, I do not have a UPS. This might be a
stretch, but I recently recieved a notice from my apartment complex's
office stating that power will be shut off this coming Tuesday because
they "are installing new electrical service to this building." I'm not
sure if this is relevant or not, but I figured it's worth a mention.

Now, the rest of the story:

The original shenanigans started last weekend. At first I though
something somehow got corrupted in Windows (perhaps due to some odd
problems with updating AVG's definitions). There was a frozen screen
similar to what I mentioned above. The reset button on my PC wouldn't
work, so I pressed the power button. I waited 30 seconds and turned on
the PC again. I got this:

Windows could not start because the following file is missing or
corrupt:
\WINDOWS\SYSTEM32\CONFIG\SYSTEM

After Googling this error, I tried the method outlined here:

"How to recover from a corrupted registry that prevents Windows XP from
starting"
http://support.microsoft.com/kb/307545

That seemed to the trick.

But the same problem came back and I later got a reboot loop. If I
remember correctly, I finally realized I needed to disable the reboot on
failure feature. I was then greeted with this BSOD while trying to boot
into Windows:

STOP: 0x0000007E (0xC0000005, 0xF73C1D7C, 0xF78D2038, 0xF78D1D34)

acpi.sys - Address F73C1D73 base at F73AE000, DateStamp 480252b1

Recommendations were to:

1. Check for adequate disk space (no problem... 80% free)
2. Check drivers
3. Check for a change to the video adapter
4. Check for a BIOS update (mine is the latest)
5. Disable BIOS memory options such as caching or shadowing (I figured
I'd hold off on this one since I wasn't too familiar with it).

At this point, since I have a BART PE disk, I figured I'd boot off of
that. The first attempt started off fine but resulted in the following
BSOD when I was using the A43 File management utility:

*** STOP: ox00000024 (0x001902FE, 0xF78F6980, 0xF78667C0, 0xF709D1E0)

*** ntfs.sys - Address F709D1E0 base at F706E000, Datestamp 41107eea

Recommendations were to:

1. Check hard drive configuration
2. Check for any updated drivers
3. Run CHKDSK /F to check for hard drive corruption

I later attempted a second boot into BART PE. Then I was greeted with:

"BIOS in this system is not fully ACPI compliant ... turn off ACPI mode
during text mode setup."

There was also this stop error:

0x000000A5 (ox00000011, 0x00000000, 0x00000000)

It was at this point I concluded the problem is not in Windows (not the
original problem at any rate) and is not from the hard drive (by the
way, later in the week, I ran SeaTools and the drive seemed to be
healthy).

By the way, I Googled the above message and learned that pressing F7 at
the moment the F6 option of offered is the way to "silently disable
ACPI."

I then attempted to use chkdsk in both the Recovery Console and BART.
FWIW, there was an orphaned file:

avgsched.log in index $I30 of file 24276

I noted this just on the off-chance that this is relevant. (Recall I
seemed to be having problems with AVG earlier.)

Anyway, once more, I couldn't boot into Windows. Rather than repeat the
method above mentioned on the Microsoft page, I decided to run a repair
install because I wasn't 100% sure this was not a software issue.

The problems seemed to finally go away. Oddly since I hadn't reverted to
IE6 before the repair install (I had been running IE7), the repair
install of course changed IE back to IE6. Interestingly, the desktop
shortcut icon had only four choices:

1. Create shortcut (yes, this was the first choice, and double-clicking
the icon created another shortcut icon on the desktop!)

2. Delete

3. Rename

4. Properties

The above is minor in the overall scheme of things, but I figured I'd
include it.

Well, I'm sure you guessed it. The BSODs returned. Here's one:

Error Signature

BCCode: 1000008e BCP1:C0000005
BCP2:8063327C BCP3:AA7D58B0
BCP4:00000000 OSver:5_1_2600
SP:2_0 Product:768_1

There was also another one with a dump:

IRQL_NOT_LESS_OR_EQUAL

STOP: ox0000000A (0x0000000F, 0x00000002, 0x00000000, 0x80505809)

Beginning dump of physical memory.
Dumping physical memory to disk :42

C:\DOCUME~1\DAVETH~1\LOCALS~1\Temp\WERaecl.dir00\Mini090908-02.dmp
sysdata.xml

At this point, whenever I would try to boot into Windows (normal or Safe
Mode), all I would get was a black screen and moveable cursor. (Actually
in Safe Mode, it wasn't *totally* black; there was the usual indication
of Safe Mode in the corners.)

I decided to run Memtest 86+ because I had a feeling this was all due to
memory problems. It ran for over seven hours (22 passes!) and no errors
were found. Then I ran the Windows Memory Diagnostic. Although no errors
(in over 40 passes) were found initially when I ran it in standard mode,
thousands were later found in Extended mode (after nearly three passes).
Sure enough, these errors were associated with the 1GB stick in slot 1.
It should be noted that the final tally was 7,871 errors for RAM stick
#1. Unfortunately, there were four errors associated with RAM stick #2.
Zero would have been much more comforting! Anyway, I removed (what I
assumed was) the faulty module and placed the other 1GB stick (which was
in slot 3) into slot 1.

Removing the "bad" RAM seemed to fix the problem. "Seemed" is the
operative word.

This takes us back to yesterday, when I figured I'd restore an image of
my hard drive I had made a few weeks ago back when my system was stable.
This wouldn't be a problem, since my data backups were current. I used
Acronis True Image. I have the ATI plugin on my BART PE disk, which
worked quite nicely (only 15 minutes to restore the image). This was
done last night. It seemed successful. However, after I went to bed, the
problem mentioned above occurred at 3:40 AM (the freeze).

I apologize for such a long post. I just wanted to make sure I didn't
leave out any useful clues.

If you have gotten this far, dear reader, I thank you from the bottom of
my heart!

 

[Dave]

I'll start at the end.

Last night I left my PC on. I was downloading the installation file for
Ultimate Boot CD for Windows. The download apparently had completed. And
then at some later time, the PC seemed to freeze. Since I had woken up
in the wee hours of the morning, I went to the PC (at about 4:15 AM) and
I noticed that the screen was frozen: not only could I not move the
cursor with my mouse, but the time display in the system tray was frozen
at 3:40 AM.

(snip)

You've got several symptoms that all seem to point to a bad hard drive.
Whatever hard drive you've got XP installed on seems to be going bad.

However, this is going to be a tricky one to narrow down, because the same
symptoms can be caused by a bad data cable (hard drive) or a bad power
supply, or a bad hard drive controller (bad mainboard).

Since you've got Acronis True image and two hard drives, I would suggest:
1) Backup ALL data files
2) Remove the XP hard drive (physically remove it!) and hard drive cable
from the system.
3) Restore XP to the "other" hard drive using Acronis True Image, and
PREFERABLY an image of your system that you created before you first noticed
ANY symptom. The reason for this is, if your hard drive was going bad, it's
possible that even your backup images might have some corrupt files in them.
So try to restore from an image made before ummm, shit started happening.
4) If you don't have an image that is older than when the symptoms started,
you could always do a full install of Windows and everything else from
scratch, but this would take a lot longer, obviously.

If that doesn't work, I'd suspect a bad power supply next. -Dave

 

[Dave]

Those are certainly possibilities. I might do just that (i.e., take the
internal hard drive with Windows out of the equation). But here is my new
working theory:

I believe the Windows Memeory Diagnostic results were a red herring. These
problems occur only if my external hard drive is plugged in (USB). Or it
might be my printer (also USB). I have disconnected both and the behavior
is totally gone. Sure, it's possible it's intermittent and it will return,
and the cause may be a hard drive going bad. But I really don't think so.
I booted off BART PE and had problems as well. As far as I understand, the
hard drive doesn't come into play. Then again, it might if for a swapfile
is created. Booting off a live Knoppix CD didn't seem to lead to problems.
Then again, perhaps if I had run it long enough...

Anyway, the first thing I will do is now is doublecheck my USB controller
drivers. Or is it possible the drivers are fine and the culprit is indeed
the external hard drive? If so, I wonder what the mechanism is. This is
new territory for me!

One last idea: it seemed that some of the crashes didn't occur until I
used my mouse -- yep... a USB mouse! I think I might be onto something...
I think I'll hunt for an older PS/2 mouse, too.

Hmmmmm...well, USB peripherals do put a load on a power supply. It could
possibly be a weak power supply, and adding too much USB stuff is causing it
to flake out? That would almost make sense, as it COULD lead to symptoms
that you wrote about. Knoppix isn't going to tax a power supply as much as
Windows XP will. If you can run for several hours with knoppix without
problems, then maybe you should look at the power supply. -Dave

 

[Paul]

I'll start at the end.

Last night I left my PC on. I was downloading the installation file for
Ultimate Boot CD for Windows. The download apparently had completed. And
then at some later time, the PC seemed to freeze. Since I had woken up
in the wee hours of the morning, I went to the PC (at about 4:15 AM) and
I noticed that the screen was frozen: not only could I not move the
cursor with my mouse, but the time display in the system tray was frozen
at 3:40 AM.

I assume something significant happened at 3:40 AM.

Everything *looked* okay. The lights and fans were working. The little
light next to where the LAN cable plugs in showed activity. Nothing was
hot. I noted the temperatures in Abit's uGuru utility, and they were
fine.

I decided to run the Windows Memory Diagnostic (just the Standard
tests). In the first pass, there were failures. I think this is
significant. (The significance will become more apparent when I discuss
the history below.) But I'm not sure if this points to memory errors or
possibly something else (e.g., something wrong with the motherboard).

FWIW, there was nothing significant in the Event Viewer.

I also ran the SeaTools hard drive diagnostic, which was negative.

Just for yucks, I disconnected the printer and external hard drive (both
USB devices). I rebooted and ran WMD again. After nearly 2 1/2 hours and
three passes in Extended mode, there were no errors!

So far, here are my ideas as to what is causing this issue:

1. Since this behavior has *not* happened with either USB device plugged
in (not yet, anyway), there might be an anomaly associated with one of
them (probably the external hard drive) that is responsible. I somehow
doubt this is the case, but I figured I'd throw it out there, anyway.

2. My wife's personal care attendant was moving stuff around the other
day, and I believe the PC (which is on the floor) was moved slightly
while it was running. Whatever happened as a result might be the cause
as these hardware problems occurred after this incident.

3. BIOS settings are not quite right or there is something wrong with
the CMOS chip or perhaps motherboard battery (which I haven't replaced
just yet). Here are the specs of my PC:

- Abit IX38 Quad GT motherboard

- Intel Core 2 Duo E8400 CPU

- Crucial Ballistix PC8500 DDR2 RAM (2 x 1GB)

- 250 GB Seagate ES.2 SATA2 hard drive (this is the one that has Windows
XP, SP2)

- 500 GB Seagate 7200.11 SATA2 hard drive (for storage)

- Lite-On 20x DL DVD +/- RW drive

- A-922 case

- 600 watt Tiger power supply

- MSI 512MB HD3870 OC PCI-E graphics card

Regarding this last item, when I originally built the system this past
April, the connection going into the card was not 100%. I still haven't
gotten a newer monitor; I am using my old Samsung CRT monitor which
requires an adapter. Once the connection was more solid, all was well
(for over 4 months). Just to be on the safe side, I checked this
connection again. I also checked that the card was seated properly.

Here are some of the BIOS settings:

Frequency: 3230 MHz
CPU Operating Speed: User Define
Multiplier Factor: 9.5
Estimated New CPU clock: 3230 MHz

This last item is in gray and has an X next to it. It seems as if this
is overclocked. If so, I wonder how this happened as I never defined
these settings. Might this be the issue?

Other gray lines with Xs:

Target CPU core voltage: 1.2250 V
DDR2 Voltage: Auto
CPU Vit Voltage: 1.10 V
MCH 1.25V Voltage: 1.25 V
ICH 1.05V Voltage: 1.05 V
ICHIO 1.5 Voltage: 1.50 V
DDR2 Reference Voltage: Default
CPU GTLREF 0 & 2
CPU GTLREF 1 & 3

Nothing seems out of the ordinary here.

Also, CPU fan speed (under Fan Speed Monitoring): 1680-1740 RPM.

In Integrated Peripherals, On Chip SATA device:

SATA mode: IDE (other choices are RAID and AHCI)

Speaking of SATA, when I installed XP, there was the option to press F6
if I needed to install SCSI or RAID drivers. I never did this because I
do not have a RAID setup. And the installation went fine. Should I have
installed SATA drivers from the get-go? Or did I do it correctly?

Also, in the options section of the BIOS is an option to load fail-safe
defaults. I haven't done this yet. Is it useful to try?

Oddly, another time I went into the uGuru utility, the values for the
CPU had changed! They were:

CPU Operating Speed: 3000 (333)
Multiplier Factor: 9
Estimated New CPU clock: 3060 MHz

So, I'm not really sure what's going on here.

Finally, since I mention ACPI below, the BIOS value for ACPI Suspend
Type is S3 (Suspend-To-RAM). The other choice is S1 (Power On - Suspend)

4. There is still a loose connection I have overlooked.

5. There might be something wrong with the motherboard.

6. There might be a problem with the power supply. Although I do have a
multimeter, I haven't used it yet. But that might be one of the next
steps.

7. A power surge. Although I have a power strip that is supposed to
protect against power surges, I do not have a UPS. This might be a
stretch, but I recently recieved a notice from my apartment complex's
office stating that power will be shut off this coming Tuesday because
they "are installing new electrical service to this building." I'm not
sure if this is relevant or not, but I figured it's worth a mention.

Now, the rest of the story:

The original shenanigans started last weekend. At first I though
something somehow got corrupted in Windows (perhaps due to some odd
problems with updating AVG's definitions). There was a frozen screen
similar to what I mentioned above. The reset button on my PC wouldn't
work, so I pressed the power button. I waited 30 seconds and turned on
the PC again. I got this:

Windows could not start because the following file is missing or
corrupt:
\WINDOWS\SYSTEM32\CONFIG\SYSTEM

After Googling this error, I tried the method outlined here:

"How to recover from a corrupted registry that prevents Windows XP from
starting"
http://support.microsoft.com/kb/307545

That seemed to the trick.

But the same problem came back and I later got a reboot loop. If I
remember correctly, I finally realized I needed to disable the reboot on
failure feature. I was then greeted with this BSOD while trying to boot
into Windows:

STOP: 0x0000007E (0xC0000005, 0xF73C1D7C, 0xF78D2038, 0xF78D1D34)

acpi.sys - Address F73C1D73 base at F73AE000, DateStamp 480252b1

Recommendations were to:

1. Check for adequate disk space (no problem... 80% free)
2. Check drivers
3. Check for a change to the video adapter
4. Check for a BIOS update (mine is the latest)
5. Disable BIOS memory options such as caching or shadowing (I figured
I'd hold off on this one since I wasn't too familiar with it).

At this point, since I have a BART PE disk, I figured I'd boot off of
that. The first attempt started off fine but resulted in the following
BSOD when I was using the A43 File management utility:

*** STOP: ox00000024 (0x001902FE, 0xF78F6980, 0xF78667C0, 0xF709D1E0)

*** ntfs.sys - Address F709D1E0 base at F706E000, Datestamp 41107eea

Recommendations were to:

1. Check hard drive configuration
2. Check for any updated drivers
3. Run CHKDSK /F to check for hard drive corruption

I later attempted a second boot into BART PE. Then I was greeted with:

"BIOS in this system is not fully ACPI compliant ... turn off ACPI mode
during text mode setup."

There was also this stop error:

0x000000A5 (ox00000011, 0x00000000, 0x00000000)

It was at this point I concluded the problem is not in Windows (not the
original problem at any rate) and is not from the hard drive (by the
way, later in the week, I ran SeaTools and the drive seemed to be
healthy).

By the way, I Googled the above message and learned that pressing F7 at
the moment the F6 option of offered is the way to "silently disable
ACPI."

I then attempted to use chkdsk in both the Recovery Console and BART.
FWIW, there was an orphaned file:

avgsched.log in index $I30 of file 24276

I noted this just on the off-chance that this is relevant. (Recall I
seemed to be having problems with AVG earlier.)

Anyway, once more, I couldn't boot into Windows. Rather than repeat the
method above mentioned on the Microsoft page, I decided to run a repair
install because I wasn't 100% sure this was not a software issue.

The problems seemed to finally go away. Oddly since I hadn't reverted to
IE6 before the repair install (I had been running IE7), the repair
install of course changed IE back to IE6. Interestingly, the desktop
shortcut icon had only four choices:

1. Create shortcut (yes, this was the first choice, and double-clicking
the icon created another shortcut icon on the desktop!)

2. Delete

3. Rename

4. Properties

The above is minor in the overall scheme of things, but I figured I'd
include it.

Well, I'm sure you guessed it. The BSODs returned. Here's one:

Error Signature

BCCode: 1000008e BCP1:C0000005
BCP2:8063327C BCP3:AA7D58B0
BCP4:00000000 OSver:5_1_2600
SP:2_0 Product:768_1

There was also another one with a dump:

IRQL_NOT_LESS_OR_EQUAL

STOP: ox0000000A (0x0000000F, 0x00000002, 0x00000000, 0x80505809)

Beginning dump of physical memory.
Dumping physical memory to disk :42

C:\DOCUME~1\DAVETH~1\LOCALS~1\Temp\WERaecl.dir00\Mini090908-02.dmp
sysdata.xml

At this point, whenever I would try to boot into Windows (normal or Safe
Mode), all I would get was a black screen and moveable cursor. (Actually
in Safe Mode, it wasn't *totally* black; there was the usual indication
of Safe Mode in the corners.)

I decided to run Memtest 86+ because I had a feeling this was all due to
memory problems. It ran for over seven hours (22 passes!) and no errors
were found. Then I ran the Windows Memory Diagnostic. Although no errors
(in over 40 passes) were found initially when I ran it in standard mode,
thousands were later found in Extended mode (after nearly three passes).
Sure enough, these errors were associated with the 1GB stick in slot 1.
It should be noted that the final tally was 7,871 errors for RAM stick
#1. Unfortunately, there were four errors associated with RAM stick #2.
Zero would have been much more comforting! Anyway, I removed (what I
assumed was) the faulty module and placed the other 1GB stick (which was
in slot 3) into slot 1.

Removing the "bad" RAM seemed to fix the problem. "Seemed" is the
operative word.

This takes us back to yesterday, when I figured I'd restore an image of
my hard drive I had made a few weeks ago back when my system was stable.
This wouldn't be a problem, since my data backups were current. I used
Acronis True Image. I have the ATI plugin on my BART PE disk, which
worked quite nicely (only 15 minutes to restore the image). This was
done last night. It seemed successful. However, after I went to bed, the
problem mentioned above occurred at 3:40 AM (the freeze).

I apologize for such a long post. I just wanted to make sure I didn't
leave out any useful clues.

If you have gotten this far, dear reader, I thank you from the bottom of
my heart!

This stuff is rated 2.2V. What voltage are you using ?

http://www.crucial.com/store/partspecs.aspx?IMODULE=BL2KIT12864AL1065

Voltage and speed go hand in hand. The RAM can run at PC2-8500 or it
can be run slower. At slower speeds, less voltage would be needed to
make it stable. You need a way to determine what voltage is being
used, even if it means setting "DDR2 Voltage: Manual" and just
trying different voltages. Depending on the conditions, you don't
need the entire 2.2V. For DDR2, 1.8V is nominal (DDR3 is 1.5V for
comparison). So the memory can be run between 1.8V and 2.2V, as
specified by the manufacturer. Select a voltage and then test.
Only use as much voltage as is needed to make the RAM stable.
Excess voltage *could* shorten the life (and it could be
worse if you have four sticks and there isn't much air movement
near them).

For testing, I use -

1) Memtest86+. I only run a pass or two. The test isn't good enough as
an acceptance test. It is intended to detect RAM that is so
decrepit that it would corrupt Windows. It is good for detecting
"stuck at" faults in a memory, but is less useful as a stress test.
Possibly some of the optional tests would give better coverage,
but who can wait that long

2) I boot a Linux LiveCD such as Knoppix or Ubuntu. Then, download
a copy of Prime95 from mersenne.org . I run the Torture Test, and
run it for up to four hours. The program will stop on the first
error it finds. No errors are acceptable. Prime95 cannot test
the area where the OS is loaded, so doesn't stress all memory.
On the other hand, Memtest86+ tests more of the memory, even
"lifting" the Memtest86+ executable out of the way. The places
Memtest86+ cannot test, are the BIOS E820 reserved locations.
Perhaps that is 1MB or less of memory. What Memtest86+ lacks
in one department, it makes up for in another (total memory tested).

So my main test, is Prime95. You can also run Prime95 in Windows,
but my assumption here, is you're tired of corrupting Windows.
The nice thing about the Linux LiveCDs, is they don't install
any software on disk, so no Windows disks get trashed while
using it. The Linux utilities and desktop, execute from the CD.

For Prime95 in Windows, assuming you somehow get booted and running
again, I like this version. This runs a thread per core, so
in your case, you should see two threads running. This'll warm
up the CPU a bit. If this is throwing errors, with your one
good stick of RAM, then it is time to bump up the voltage.
You don't want to try too radical a change to your BIOS
settings, because Windows could get corrupted on the next
boot. I really prefer Linux, if it looks like my hardware
is in bad health.

http://www.mersenne.org/gimps/p95v256.zip

Another reason for RAM trouble, is a mismatch between RAM
timings and RAM clock speed. When amateur overclockers play
with a machine, then can unintentionally increase RAM clock
speed. When you do that, the BIOS may require manual correction
of the timing numbers (slacken CAS, Trcd, and so on). The BIOS
doesn't correct every sin the user commits. It is amazing what
some people get away with, when the RAM should be throwing a
fit because of the abuse

Another good tool is CPUZ, because it gives a snapshot of
what the BIOS has done to you. For example, my motherboard
lies about RAM timing, and CPUZ helped me detect the problem.
This runs in Windows.

http://www.cpuid.com/cpuz.php

HTH,
Paul

 

[Dave]

No offense, but wouldn't it be easier to test for a bad power supply
first?

Yes, if he happens to have a good spare on hand. But if his symptoms are
caused by a power supply, then that specific power supply will test FINE
with a multimeter. So again, testing for a bad power supply is a good idea,
if by "test", you mean that you are going to replace it with a good one.

So many people suggest testing power supplies with voltmeters or multimeters
and I totally disagree with this. It is a waste of time. If your power
supply is so bad that it tests bad on a multimeter, then the multimeter
testing will be redundant. That is, there will be no doubt in your mind
that the power supply IS bad, and therefore the multimeter will just give
you more information about how bad it is.

Testing a power supply with a voltmeter/multimeter, is about like using an
air pressure guage on a tire that has a huge hole all the way through the
sidewall. Do you really need to confirm that there is no air pressure in
the tire, after you are 100% certain that the tire won't hold air at all???

In the same manner, if you are not SURE that the power supply is bad, then
pulling out the multimeter will just show voltages that are within spec. at
the moment you test it. But that bit of information isn't helpful at all,
beyond telling you that the power supply isn't totally dead, yet. You
specifically do NOT know if the power supply is good, just because the
voltages read within specs for a few seconds. -Dave

 

[Sjouke Burry]

Paul:
To stray just a little off-subject: After reading your testing method
using a Linux boot CD I downloaded Ubuntu AND Knoppix iso. I burned a
CD with each. When I boot with them they come up with errors. Ubuntu
says the CD can't be verified (no, it didn't say that, can't remember
their wording, but that was the meaning). I burned it onto a DVD, same
thing. Knoppix says 'can't find a knoppix file system'. The Ubuntu
(desktop-i386) download passed a hash check that they suggest that
you do so it was good that far. Do I have a bad burner or am I doing
something else stupid?

My knoppix cd worked first time, recognizing soundcard, ethernet card,
,cd and network.
No tweaks needed.

 

[Paul]

Paul:
To stray just a little off-subject: After reading your testing method
using a Linux boot CD I downloaded Ubuntu AND Knoppix iso. I burned a
CD with each. When I boot with them they come up with errors. Ubuntu
says the CD can't be verified (no, it didn't say that, can't remember
their wording, but that was the meaning). I burned it onto a DVD, same
thing. Knoppix says 'can't find a knoppix file system'. The Ubuntu
(desktop-i386) download passed a hash check that they suggest that
you do so it was good that far. Do I have a bad burner or am I doing
something else stupid?

The distros are in ISO9660 format. You need a program that
handles ISO9660 images, and transfers them to the CD. I used
Nero for the job. In addition, you need a burner that can
handle a 700MB burn. The first time I tried to prepare my
Linux CD, I discovered to my horror that the burner I was
using, couldn't actually do a burn that big. I had to buy
a new burner to handle the 700MB file.

When I burn with Nero, there is a scanning function for
verifying the media afterwards. I used that function to
"dial in" my spindle of media. I discovered, that the
burn would only be good at 4X, and as long as I don't
try to burn faster than that, the discs are perfect.

My Knoppix CD does work with Windows. If the CD is inserted,
there is an autorun file on the CD, that causes a browser to
open and a Knoppix web page to appear. If I explore the CD,
I can see a "boot" and a "knoppix" folder, and in the latter
folder, there is a 700MB file visible. I expect that is
the compressed file system that Knoppix eventually
mounts. (The 700MB file is read when needed at runtime,
and is compressed. So the OS decompresses the necessary pieces
of it, to launch applications and the like.)

For Knoppix, you should study the "cheat codes" web page,
as it has a number of interesting options. These are
useful if you're getting as far as the "boot prompt".
If you aren't even getting that far, then you'd probably
want to try the CD in some other computer first.
For example "knoppix testcd" at boot time, checks the
CD contents for you.

http://knoppix.net/wiki/Cheat_Codes

Knoppix even includes a copy of memtest, and at the boot
prompt you can type "memtest" and thus the Knoppix CD
doubles as your memtest boot media.

Paul

 

[Daave]

Hmmmmm...well, USB peripherals do put a load on a power supply. It
could possibly be a weak power supply, and adding too much USB stuff
is causing it to flake out? That would almost make sense, as it
COULD lead to symptoms that you wrote about. Knoppix isn't going to
tax a power supply as much as Windows XP will. If you can run for
several hours with knoppix without problems, then maybe you should
look at the power supply.

Thanks for the feedback, Paul.

Per Paul's suggestion, I bumped up the DDR2 voltage to 2.1V. I
immediately ran Windows Memory Diagnostic (booting off floppy). Within a
half hour or so (during the first pass), there were nearly 100 errors
*and* WMD froze (actually, the word "trap" appeared). I then decided to
load "fail-safe" defaults to see what would happen. Right off the bat,
WMD showed errors -- both in Slot A2! (instead of the usual Slot A0).
Also, there was another "trap."

I reverted to my original settings and rebooted. Then something new
happened: an Award BootBlock message showed up: BIOS ROM checksum error.
A hard boot fixed that problem. I wonder if I should flash the BIOS at
this point.

But I do believe a flaky power supply might be the cause of all these
shenanigans, so that's the next thing I will look at. I am also going to
explorer BIOS settings.

As I mentioned in another post, there have been similar reports from
others:

http://forum.uabit.com/showthread.php?t=136965

If the power supply turns out to be fine, my next guess is that *both*
sticks of RAM are bad.

 

[Daave]

Thanks for the feedback, Paul.

I mean Dave! (you'd think I'd remember that name...)

 

[Paul]

I wonder if I should flash the BIOS at this point.

Not while a computer is demonstrating how flaky it can be.
Imagine what happens if there is a RAM error during the flash.

Paul

 

[Paul]

I agree.

It's a Tiger USA ATX-600W P4. It's under warranty, but I'd rather just
get a high quality PSU. I'll have the store where I purchased this
arrange for an exchange.

Any recommendations? I understand Antec and OCZ are popular.

Here again is what I have:

- Abit IX38 Quad GT motherboard

- Intel Core 2 Duo E8400 CPU

- Crucial Ballistix PC8500 DDR2 RAM (2 x 1GB)

- 250 GB Seagate ES.2 SATA2 hard drive (this is the one that has Windows
XP, SP2)

- 500 GB Seagate 7200.11 SATA2 hard drive (for storage)

- Lite-On 20x DL DVD +/- RW drive

- A-922 case

- 600 watt Tiger power supply

- MSI 512MB HD3870 OC PCI-E graphics card

What is the rule of thumb for how many watts one needs? I doubt I need
600 watts.

If you aren't overclocking the video card, it draws 81.1 watts. The E8400
is 65W. Motherboard plus RAM is around 50W allocation. Two hard drives
total 26W. Optical 25W while spinning, maybe 5-7W otherwise. So 250W
if even that much. If you're doing some overclocking, it could be
a bit higher. By the time you buy a supply with a PCI Express power
connector, chances are you've got enough power. (In this article,
they volt mod and overclock a 3870, presumably to see how much
power they could draw )

http://www.xbitlabs.com/articles/video/display/radeon-hd3870-voltmod_4.html

In terms of amperes (like the limits shown on the supply label), you
should work out 12V1 (motherboard) and 12V2 (CPU) current values.
65W/12V * (1/0.90) = 6 amps for your CPU from 12V2. The 12V1
rail provides the rest. 6.75A for video card. 0.5A fans.
0.6A*2 for HDD. 1.5A for ODD. About 10A for 12V1. I'd probably
aim for a supply with at least 13A on both 12V1 and 12V2 (as
chances are, they have close to the same rating).

This would be about as low as I'd go (the product name "Stallion" isn't
helping matters - makes me think of the glue factory). The 400W max
would seem to be enough, and the 12V1 meets the load with a little margin,
but the 5V fails another criterion. I like to see 20A offered on 3.3V
and on 5V, just in case. This is because it isn't possible to work out
exact numbers for those rails, and having at least 20A will likely
keep you out of trouble. This supply has one PCI Express power
connector. $37

+3.3V @ 20A, +5V @ 14A, +12V1 @ 14A, +12V2 @ 13A, -12V @ 0.3A, +5VSB @ 2.5A, 400W max
http://www.newegg.com/Product/Product.aspx?Item=N82E16817182074

A quality brand name, with capacities a bit better than that, should be
enough. For example, this one is an 80% efficiency model, which means
less waste heat comes out of the supply itself.

+3.3V @ 20A, +5V @ 20A, +12V1 @ 17A, +12V2 @ 17A, -12V @ 0.8A, +5VSB @ 2.5A
430W max.

http://www.newegg.com/Product/ProductReview.aspx?Item=N82E16817371006

Paul

 

[Daave]

If you aren't overclocking the video card, it draws 81.1 watts. The
E8400 is 65W. Motherboard plus RAM is around 50W allocation. Two hard
drives
total 26W. Optical 25W while spinning, maybe 5-7W otherwise. So 250W
if even that much. If you're doing some overclocking, it could be
a bit higher. By the time you buy a supply with a PCI Express power
connector, chances are you've got enough power. (In this article,
they volt mod and overclock a 3870, presumably to see how much
power they could draw )

http://www.xbitlabs.com/articles/video/display/radeon-hd3870-voltmod_4.html

In terms of amperes (like the limits shown on the supply label), you
should work out 12V1 (motherboard) and 12V2 (CPU) current values.
65W/12V * (1/0.90) = 6 amps for your CPU from 12V2. The 12V1
rail provides the rest. 6.75A for video card. 0.5A fans.
0.6A*2 for HDD. 1.5A for ODD. About 10A for 12V1. I'd probably
aim for a supply with at least 13A on both 12V1 and 12V2 (as
chances are, they have close to the same rating).

This would be about as low as I'd go (the product name "Stallion"
isn't helping matters - makes me think of the glue factory). The 400W
max
would seem to be enough, and the 12V1 meets the load with a little
margin, but the 5V fails another criterion. I like to see 20A offered
on 3.3V
and on 5V, just in case. This is because it isn't possible to work out
exact numbers for those rails, and having at least 20A will likely
keep you out of trouble. This supply has one PCI Express power
connector. $37

+3.3V @ 20A, +5V @ 14A, +12V1 @ 14A, +12V2 @ 13A, -12V @ 0.3A, +5VSB
@ 2.5A, 400W max
http://www.newegg.com/Product/Product.aspx?Item=N82E16817182074
A quality brand name, with capacities a bit better than that, should
be enough. For example, this one is an 80% efficiency model, which
means
less waste heat comes out of the supply itself.

+3.3V @ 20A, +5V @ 20A, +12V1 @ 17A, +12V2 @ 17A, -12V @ 0.8A, +5VSB
@ 2.5A 430W max.

http://www.newegg.com/Product/ProductReview.aspx?Item=N82E16817371006

Thanks a million for the crash course, Paul. It makes a lot more sense
now.

I guess New Egg's calculator isn't very accurate. Their recommended
wattage is 546. I wonder why there is such a discrepancy.

 

[Paul]

Thanks a million for the crash course, Paul. It makes a lot more sense
now.

I guess New Egg's calculator isn't very accurate. Their recommended
wattage is 546. I wonder why there is such a discrepancy.

There was only one power supply calculator that even came
close. Takaman.jp used to have a calculator, and it delivered
results in a spreadsheet style. It showed the power component
for 3.3V, 5V, and 12V, and I think it may have also mentioned
where the estimates were coming from. At one time, you could
buy a tech report, with measured power number for the video
cards, and that is where they got their video card numbers.

With Takaman.jp gone, Xbitlabs provides measured values for
video cards, and you can use TDP for the processor itself.
I've measured the current flow on all rails, on a couple
motherboards here, which influenced my recommendation. But
since I don't upgrade that regularly, I haven't had a
chance to check any recent boards with my clamp-on DC
ammeter.

And as a sanity check, someone used a Kill-O-Watt meter
on their PC, I used my estimation method, and I still
ended up being on the high side of the truth (even taking
power factor phase angle on the AC into account).

But some of the currently available web sites, give numbers
about double what I give. I've seen sites that estimate
a stick of RAM at 25W, and if a memory DIMM actually
used that much, you'd get a serious burn if you
touched one Kingston has datasheets for RAM, and
some of those will show a 3W number for the industry
defacto standard cycle mix. (RAM power depends strongly
on cycle type, so if you put a string of the right type
of cycles back to back, the stick would get hot. But
real processors don't use memory that way, and so there
is an assumed cycle mix that is used when quoting RAM
power consumption. 3W to 5W should be plenty per stick.)

I now bundle the chipset power (maybe 25W-30W on some of
the sets) plus 4*3W, and round it to 50W and call that
enough for a motherboard plus RAM. It isn't worth the
time to try to refine those numbers any further.

For hard drives and optical drives, you can get some
representative numbers from product documentation.
My optical uses 12V @ 1A for the motor, when media is
present, while the nameplate rating is 12V @ 1.5A.

Paul

 

[Daave]

This stuff is rated 2.2V. What voltage are you using ?

http://www.crucial.com/store/partspecs.aspx?IMODULE=BL2KIT12864AL1065

The above dead link used to reference my RAM:

Crucial Ballistix PC8500 DDR2 RAM (2 x 1GB)
BL2KIT12864AA1065

See:

http://www.newegg.com/Product/Product.aspx?Item=N82E16820148069

(Item#:N82E16820148069) Note that the page says:
"Deactivated Item!"

Okay, I believe I now know what has been going on. Apparently my RAM
doesn't play well with my motherboard (Abit IX38 Quad GT)! When I was
selecting components for my new PC build several months ago, apparently
the Newegg configuration tool didn't get the memo that this RAM module
is not compatible with the Intel X38 chipset. The problems I have been
experiencing (blue screens while running Windows or even my BART PE CD
and RAM errors while running Windows Memory Diagnostic -- but no errors
at all while running Memtest86+ or Prime95!) were intermittent and
tended to occur when the PC was on for an extended period of time
(normally I power it down at nighttime) *and* when my USB external hard
drive was plugged in.

I intend to exchange the two modules for a more compatible pair. Any
suggestions? If anyone has the same mobo I have, I would love to get
your input.

Voltage and speed go hand in hand. The RAM can run at PC2-8500 or it
can be run slower. At slower speeds, less voltage would be needed to
make it stable. You need a way to determine what voltage is being
used, even if it means setting "DDR2 Voltage: Manual" and just
trying different voltages. Depending on the conditions, you don't
need the entire 2.2V. For DDR2, 1.8V is nominal (DDR3 is 1.5V for
comparison). So the memory can be run between 1.8V and 2.2V, as
specified by the manufacturer. Select a voltage and then test.
Only use as much voltage as is needed to make the RAM stable.
Excess voltage *could* shorten the life (and it could be
worse if you have four sticks and there isn't much air movement
near them).

In another post I indicated that the BIOS setting for Voltage Control
(and thus DDR2 voltage) was Auto (detect). When using the BIOS utility,
I noted that the level was 1.97V, which was right in the middle of the
1.8 to 2.2V range. When I tried 2.1, the problem just worsened, so I
never even tried the "rated" 2.2V. (However, the JEDEC #3 figure is
1.8V!)

Anywho, I was thinking of trying one final thing before exchanging the
RAM: using manual settings. Currently, the value for DRAM Timing
Selectable is "By SPD." Although TCL, TRCD, and TRP are both 5, there is
a discrepancy with TRAS! The SPD (JEDEC #3) value is 18, but the spec
sheet says 15! Is it worth playing around with these values? I somehow
doubt it, but if someone has been successful, I'm all ears. Here is what
I currently have in my BIOS:

CPU OPERATING SPEED................ 3,000 (333)
EXTERNAL CLOCK..................... 340 MHz
MULTIPLIER FACTOR.................. 9.0
# ESTIMATED NEW CPU CLOCK.......... 3060 MHz
DRAM SPEED (CPU : RAM)....... Default (DDR2-816)
PCI EXPRESS CLOCK.................. 100 MHz

VOLTAGE CONTROL: Auto Detect

CPU CORE VOLTAGE................... 1.2250 V
DDR2 VOLTAGE....................... Auto
CPU VTT 1.2V VOLTAGE............... 1.10 V (why isn't his 1.2 V?)
MCH 1.25V VOLTAGE.................. 1.25 V
ICH 1.05V VOLTAGE.................. 1.05 V
ICHIO 1.5V VOLTAGE................. 1.50 V
DDR2 REFERENCE VOLTAGE............. Default
CPU GTLREF 0&2..................... 67%
CPU GTLREF 1&3..................... 67%

ADVANCED CHIPSET FEATURES:

DRAM TIMING SELECTABLE............. By SPD
CAS LATENCY TIME (TCL)............. Auto
RAS# TO CAS# DELAY (TRCD).......... Auto
RAS# PRECHARGE (TRP)............... Auto
PRECHARGE DELAY (TRAS)............. Auto
REFRESH CYCLE TIME (TRFC).......... Auto
WRITE RECOVERY TIME (TWR).......... Auto
WRITE TO READ DELAY (TWTR)......... Auto
ACT TO ACT TIME (TRRD)............. Auto
READ TO PRECHARGE (TRTP)........... Auto
COMMAND RATE....................... Auto
PEG FORCE X1....................... Disabled
INIT DISPLAY FIRST................. PCI slot

ADVANCED BIOS FEATURES / CPU FEATURES:

THERMAL CONTROL.................... Enabled
LIMIT CPUID MAXVAL................. Disabled
C1E FUNCTION....................... Auto
EXECUTE DISABLE BIT................ Enabled
VIRTUALISATION TECHNOLOGY.......... Enabled
EIST FUNCTION...................... Enabled

For testing, I use -

1) Memtest86+. I only run a pass or two. The test isn't good enough as
an acceptance test. It is intended to detect RAM that is so
decrepit that it would corrupt Windows. It is good for detecting
"stuck at" faults in a memory, but is less useful as a stress test.
Possibly some of the optional tests would give better coverage,
but who can wait that long

As I mentioned earlier, this program (which ran several times for 16
hours a pop) did not show any errors whatsoever. Neither did Prime95's
torture test. Windows Memory Diagnostic's extended test was the only one
that indicated memory errors.

So my main test, is Prime95. You can also run Prime95 in Windows,
but my assumption here, is you're tired of corrupting Windows.
The nice thing about the Linux LiveCDs, is they don't install
any software on disk, so no Windows disks get trashed while
using it. The Linux utilities and desktop, execute from the CD.

I wound up using UBCD4Win, which has Prime95.

Another reason for RAM trouble, is a mismatch between RAM
timings and RAM clock speed. When amateur overclockers play
with a machine, then can unintentionally increase RAM clock
speed. When you do that, the BIOS may require manual correction
of the timing numbers (slacken CAS, Trcd, and so on). The BIOS
doesn't correct every sin the user commits. It is amazing what
some people get away with, when the RAM should be throwing a
fit because of the abuse

And a good reason, that. That's why I included my figures above.

Thanks in advance to Paul and anyone else who has helpful advice.

 

[Daave]

The above dead link used to reference my RAM:

Crucial Ballistix PC8500 DDR2 RAM (2 x 1GB)
BL2KIT12864AA1065

See:

http://www.newegg.com/Product/Product.aspx?Item=N82E16820148069

(Item#:N82E16820148069) Note that the page says:
"Deactivated Item!"

Good grief! Now, I just learned that my mobo (ABIT IX38 Quad GT) has
become a deactivated item too!

http://www.newegg.com/Product/Product.aspx?Item=N82E16813127033

Do I know how to pick 'em or what?

 

[Paul]

Good grief! Now, I just learned that my mobo (ABIT IX38 Quad GT) has
become a deactivated item too!

http://www.newegg.com/Product/Product.aspx?Item=N82E16813127033

Do I know how to pick 'em or what?

Abit will stop making motherboards at the end of the year.
They're moving to other consumer items, presumably ones
with more profit margin, than trying to beat the likes
of Asus or Gigabyte. Over in the Abit newsgroup, they're saying
their good-byes.

http://en.wikipedia.org/wiki/Universal_abit

"By 2009, Abit will no longer sell motherboards and may focus
on CE field. Warranties will still be handled by Abit for
another 3 years or so."

The reviews for the Ballistix on Newegg look pretty bad.
A good reason to discontinue the product, if every
customer has to do RMAs.

Memtest86+ has some optional tests, and one of them is a
refresh test. The memory is loaded with a pattern,
and left for a while, and then read out. That tests to
see if the DRAM capacitors hold their value properly,
with the refresh timer constant the user has selected.
Normally, memory holds data much longer than needed
(i.e. the default refresh setting causes refresh to
occur, way before any cells have discharged). And that
is why the test is not a standard one. You can try that
one if you like. But based on the Newegg customer reviews
for the memory, my guess would be the memory will croak
some time soon.

I think the Newegg reviews have some merit, at detecting
products to avoid, and are as good a recommendation as any.
The last memory I bought, I looked at the alternatives, and
checked for failures in the Newegg reviews before buying.

Paul

 

[Paul]

Chances are I'll exchange the RAM *very* soon. However, I do want to
exhaust the other possibilities, namely BIOS settings.

So far, I have manually set the RAM timings to 5-5-5-15 and the RAM
voltage to 2.2V.

The BSODs often have this as an underlying thread: "Disable BIOS memory
options such as caching or shadowing." Is this a red herring, or is
there something to it? If there's something to it, how is this done? I'm
pretty sure that pressing F7 during bootup accomplishes this, but is
there a way to make it permanent? In my BIOS's Power Management Setup,
the first item is ACPI Suspend Type. The normal value is S3
(Suspend-To-RAM). The other choice is S1 (Power On-Suspend). Is that it?
Or am I barking up the wrong tree? Also, if I don't change the setting,
I get the option of enabling (this is the default) or disabling "Resume
by USB from S3." The only reason I bring this up is these problems
(BSODs and memory test errors) only occur if my external USB hard drive
is plugged in.

There are a number of shadowing/caching schemes, that involve copying things
from a slow resource, into a fixed RAM area. Some of them are mentioned here.
I believe it is even possible, for a motherboard to copy the entire contents
of the BIOS EEPROM into RAM, for faster execution of BIOS code. I doubt I
could do a decent job of explaining all of the features available to do
stuff like that. This page has some skeleton explanations (at one time,
the explanations included some technical content, but not all of that
technical content was correct.)

http://www.techarp.com/freebog.aspx

ACPI Suspend Type, in the BIOS, is information passed in an ACPI table
to the OS at boot time. It tells the OS that the hardware supports S3.
When you put the computer to sleep, the current session is stored in RAM,
and when the fans stop spinning, there is still the +5VSB power rail
operating in the computer. The +5VSB powers the RAM refresh circuit, and
also delivers power to the RAM slots. It is by that means, that the session
is not lost. Unlike hibernation (S4), the RAM contents continue to be valid until
the next time the computer is awakened. In hibernation, the RAM contents are
dumped to disk, and restored from disk when the computer next awakes. The
difference is the speed of recovery, but the difference is also in terms
of the security of the session information. If the AC power goes off while
you're in S3, then the current session would be lost.

Neither of those features seems particularly associated with USB.

How can hardware affect the RAM ? Say a device uses DMA, and the DMA
pointer value is bad, and points to a chunk of RAM not allocated for
that purpose. Then all manner of havoc could result. Other than that,
the USB driver itself could be stored in an area of RAM that is corrupted.

In terms of low level issues, I've heard of a few, but don't know the details.

1) A while back, memtest used to be affected on some boards, by the
setting of "USB legacy support". The memtest would report errors, until
that setting was changed.

2) Some motherboards have had a BIOS design problem, where there is a memory
map overlap between some USB related resources, and 4GB memory configurations.
The BIOS puts up an error message about "USB overcurrent", and presumably
some info that looks like an overcurrent indication, is being read from the
wrong location. Since your memory config of 2x1GB is not an excessive one,
I doubt a resource planning problem is at the root of the problem. The
USB overcurrent thing, was on systems with 4GB of RAM (older motherboards,
not recent ones).

In the case of the first one, I don't know right off hand, how the two
things could affect one another. Maybe if you could trace down what the
issue was there, you may discover how the two are related.

http://groups.google.ca/group/alt.sys.pc-clone.dell/msg/bdf2b7e9b4b08207?dmode=source

http://www.techarp.com/showFreeBOG.aspx?lang=0&bogno=342

Your issue could have something to do with how the BIOS has set up the
resources for the motherboard. But I doubt I could figure it out, even
if I was sitting in front of the machine. I don't see any info on the
Uabit site (such as BIOS release notes), to show that they know about
any such issue.

Paul