Reproducing the error "Windows Registry Recovery..."

I have been reproducing lots of errors here lately, but this one
is notable, especially since it is serious (according to my
research).

Windows Registry Recovery
One of the files containing the system's registry data had to be
recovered by use of a log or alternate copy. The recovery was
successful.

The underlying problem is unknown (currently suspecting the
motherboard), it occurs on this system.

Windows XP SP3
Q9550, Intel CPU
GA-EP45-UD3L, Gigabyte motherboard

These actions reproduce the Windows Registry Recovery error.

.... delete this entry, Device Manager -- System Devices -- PCI bus

.... wait for a few minutes for the hard drive activity to stop

.... the mouse and keyboard stop working, so either shut down or
hit the reset button, and then restart

.... at the Windows logon screen (at this point, things might be
different if you do not use the Windows logon screen), just sit
there and wait until hard drive activity stops again, it might be
a few minutes or longer

.... instead of continuing into Windows, restart through the
shutdown dialog (this step might not actually be necessary, I
forgot to try)

.... when I get back into Windows, the Windows Registry Recovery
error reappears

There might be other ways to reproduce that error.

After all of the software possibilities are eliminated, many seem
to think the Windows Registry Recovery error is either RAM or hard
drive. In my case, it seems to be the motherboard. But I wonder
how much confidence (or not) in a hardware error since the error
can be reproduced at will. Obviously is not just on startup, since
the error does not occur every boot.

John Doe wrote:
I have been reproducing lots of errors here lately, but this one
is notable, especially since it is serious (according to my
research).

Windows Registry Recovery
One of the files containing the system's registry data had to be
recovered by use of a log or alternate copy. The recovery was
successful.

The underlying problem is unknown (currently suspecting the
motherboard), it occurs on this system.

Windows XP SP3
Q9550, Intel CPU
GA-EP45-UD3L, Gigabyte motherboard

These actions reproduce the Windows Registry Recovery error.

... delete this entry, Device Manager -- System Devices -- PCI bus

... wait for a few minutes for the hard drive activity to stop

... the mouse and keyboard stop working, so either shut down or
hit the reset button, and then restart

... at the Windows logon screen (at this point, things might be
different if you do not use the Windows logon screen), just sit
there and wait until hard drive activity stops again, it might be
a few minutes or longer

... instead of continuing into Windows, restart through the
shutdown dialog (this step might not actually be necessary, I
forgot to try)

... when I get back into Windows, the Windows Registry Recovery
error reappears

There might be other ways to reproduce that error.

After all of the software possibilities are eliminated, many seem
to think the Windows Registry Recovery error is either RAM or hard
drive. In my case, it seems to be the motherboard. But I wonder
how much confidence (or not) in a hardware error since the error
can be reproduced at will. Obviously is not just on startup, since
the error does not occur every boot.

Work through some of the suggestions here ? This thread has
more than one victim, and more than one successful resolution.
It could be bad RAM, but it could also be one of the registry
files has a structural error inside, and that is causing the
problem. There is one suggestion in this thread, that permissions
on the registry file could be an issue, but then, how would the
system be able to update the registry, if that was the case ?

"WINXP registry restore error on startup"
http://www.annoyances.org/exec/forum/winxp/t1048048020

If you use the built-in regedit, and do a search from the
beginning, looking for a non-existent quantity, does the
search go all the way to the end without incident ? I doubt
that is a really good test, because the registry is probably
parse ahead of time, and the search might not be stressing
or checking anything. The notice of trusting a "registry
cleaner" to check the registry, kinda scares me, because
there is no way of knowing whether they're reporting
real problems or not.

You could also research, how to pull registry files from
one OS install, over to another computer. I understand you
can open a hive on another computer. Perhaps such a test
case, would be a way of double checking whether the registry
is really good or not ?

Paul

John Doe wrote:


... remove the fan that was aimed at the south bridge chip and
notice that the errors occur much more frequently

... put the fan very close and notice that the errors disappear

Now the question is how long the thing will last after I put some
serious fannage on it. Not absolutely sure it is the south bridge
chip, but I guess it would be the most suspect chip in the area.

I must have rebooted between 50 and 100 times this evening.

Older Intel Southbridges ran so cool, they didn't need a heatsink.
The newer ones have more high speed stuff in them, like SATA II
and PCI Express x1 interfaces, and run a bit hotter.

I'd start by checking the Southbridge heatsink is intact, and
looks like it is in good contact. The manufacturer might use
a thermal pad, rather than thermal grease. In some cases,
you'll find an air gap between the chip and heatsink (if the
plastic pins are loose).

Or, in the case of heatpipe equipped cooling assemblies,
a leaky heatpipe running from the Southbridge to the main
heatsink block on the Northbridge, isn't working. If the tiny
bit of working fluid evaporates on a single heatpipe system,
then the chip will receive poor cooling. Some of the
heatpipe makers weren't very good at sealing them. That's
why I feel a bit better, when I see multiple heatpipes
on cooling assemblies (redundancy to leak failures).

I'd also be using alternate test environments, instead of
just Windows. Perhaps the computer will also overheat, while
you're running a Seagate Seatools disk diagnostic program ?

I don't really know the best test recipe from a Linux LiveCD,
except to perhaps copy a very large file from one disk to
another, then run MD5SUM and verify the checksums of the
two files are the same. That might tell you if disk errors
were involved.

Paul

Paul nospam needed.com wrote:

Older Intel Southbridges ran so cool, they didn't need a
heatsink. The newer ones have more high speed stuff in them,
like SATA II and PCI Express x1 interfaces, and run a bit
hotter.

How about onboard LAN stuff? Mainly curious.

I'd start by checking the Southbridge heatsink is intact, and
looks like it is in good contact. The manufacturer might use a
thermal pad, rather than thermal grease. In some cases, you'll
find an air gap between the chip and heatsink (if the plastic
pins are loose).

My GA-EP45-UD3L southbridge (the small chip) heatsink has always
been too hot to hold my finger on. The thermal pad is held in
place by light-duty springs, definitely applying some pressure.
Strangely, the thermal pad looked like it had never been used.
There was no residue on the chip and the pad was not melted or
deformed.

The thermal pad was removed and both surfaces were squeaky
cleaned. Quality superglue was applied to the southbridge chip and
the heatsink was plugged back into place. Firm pressure held
them together for about 5 minutes.

So here is the test of superglue heat transfer (I thought). Put
everything back together and fired up Windows. Ran the
aforementioned test and (@#$!) got plenty of Windows Registry
Recovery errors on restarts. Hoping to prove it was not the
superglue, I did a reinstallation of Windows, so far unable to
bring up the Windows Registry Recovery error dialog. The chip
appears to be transferring heat to the heatsink very well.

I'd also be using alternate test environments, instead of just
Windows. Perhaps the computer will also overheat, while you're
running a Seagate Seatools disk diagnostic program ?

FWIW. Installing Windows XP seems to do pretty well for heating up
the system. But most people would not be able to do that unless
they had already given up.
--








I don't really know the best test recipe from a Linux LiveCD,
except to perhaps copy a very large file from one disk to
another, then run MD5SUM and verify the checksums of the
two files are the same. That might tell you if disk errors
were involved.

Paul

John Doe wrote:
Paul nospam needed.com wrote:

Older Intel Southbridges ran so cool, they didn't need a
heatsink. The newer ones have more high speed stuff in them,
like SATA II and PCI Express x1 interfaces, and run a bit
hotter.

How about onboard LAN stuff? Mainly curious.

I'd start by checking the Southbridge heatsink is intact, and
looks like it is in good contact. The manufacturer might use a
thermal pad, rather than thermal grease. In some cases, you'll
find an air gap between the chip and heatsink (if the plastic
pins are loose).

My GA-EP45-UD3L southbridge (the small chip) heatsink has always
been too hot to hold my finger on. The thermal pad is held in
place by light-duty springs, definitely applying some pressure.
Strangely, the thermal pad looked like it had never been used.
There was no residue on the chip and the pad was not melted or
deformed.

The thermal pad was removed and both surfaces were squeaky
cleaned. Quality superglue was applied to the southbridge chip and
the heatsink was plugged back into place. Firm pressure held
them together for about 5 minutes.

So here is the test of superglue heat transfer (I thought). Put
everything back together and fired up Windows. Ran the
aforementioned test and (@#$!) got plenty of Windows Registry
Recovery errors on restarts. Hoping to prove it was not the
superglue, I did a reinstallation of Windows, so far unable to
bring up the Windows Registry Recovery error dialog. The chip
appears to be transferring heat to the heatsink very well.

I'd also be using alternate test environments, instead of just
Windows. Perhaps the computer will also overheat, while you're
running a Seagate Seatools disk diagnostic program ?

FWIW. Installing Windows XP seems to do pretty well for heating up
the system. But most people would not be able to do that unless
they had already given up.

Hmmm. If you can't keep a finger on the Southbridge heatsink,
that means it is 65C+. Allowing for delta_T to the die, you could
be around 85 to 90C at the silicon die. Something is not right there.
Generally, if you burn yourself on something on your motherboard,
you need to "fix it".

Does your motherboard have any "boost voltage" settings for the
chipset ? It sounds like perhaps the core voltage on the
Southbridge is wrong. (In the manual, I see an "ICH I/O" and
"ICH Core" setting in the MB Intelligent Tweaker screen.)

The ICH10 datasheet has these values, so it takes a bunch of
different power inputs. I've marked the entries with arrows,
that I think correspond to the settings in the BIOS. So the
values shown in the manual (1.5V and 1.1V) could very well
be the correct stock settings.

"Intel ICH10 Absolute Maximum Ratings

1.1 V Supply Voltage with respect to VSS -0.5 to 2.1 V ---
1.25 V Supply Voltage with respect to VSS -0.5V to 2.1V
1.5 V Supply Voltage with respect to VSS -0.5 to 2.1 V ---
3.3 V Supply Voltage with respect to VSS -0.5 to 4.6 V
5.0 V Supply Voltage with respect to VSS -0.5 to 5.5 V
V_CPU_IO Supply Voltage with respect to VSS -0.5 to 2.1 V"

"Vcc1_1 1.1 V supply for core well logic.
Vcc1_5_A 1.5 V supply for Logic and I/O."

Instead of Superglue, you should have used regular thermal paste,
and found another means of holding the heatsink to the chip. The
nice thing about thermal paste, is it doesn't need as much
"normal force" or "pressing downward force", to make the grease
a good conductor of heat. On chips which have bare die, you
have to be careful to use just enough grease, to allow the silicon
die to be covered when pressure is applied to the heatsink. If
the chip has a heat spreader lid on top, then you can be a
bit sloppier with the grease.

So you have two ways to attack the stability issue. Either
reduce the heat coming from the chip (turn down the core
voltage), or improve the cooling on the chip. I'm going to
see if I can find a picture of your motherboard, to see
what kind of heatsink is on it. If you use a heatsink with
more cooling fin area, then point a fan right at the
fins (impingement cooling), that is about the best you
can do on the cooling front. You might do even better
with some kind of chilled liquid cooling system, but
for the price, buying another motherboard would be
cheaper.

OK, I can see the heatsink here.

http://images17.newegg.com/is/image/newegg/13-128-372-S03?$S640W$

You may be able to replace that with a Swiftech MCX159CU.

http://www.swiftnets.com/products/mcx159-CU.asp

"Brackets allow adjustments as follows:

center to center distance between mounting holes ranges
from 2.05" (52.1mm) to 2.41" (61.4mm)."

You measure the distance from the center of one pushpin to
the other, to see if the mounting scheme would work.

That isn't the absolute best one you can use. The problem
with a lot of those after-market ones, is they don't stay in
production forever.

OK, this is another style of cooler you could look at.
This uses a heatpipe, which means more of the heat
makes it into the fins. You can point a fan at one
of these.

http://www.newegg.com/Product/Produc...82E16835233021

Thermalright makes some like that too. I think the Xigmatek is
a copy of the Thermalright. Thermalright probably made the first
ones of that style.

http://www.heatsinkfactory.com/therm...et-cooler.html

With those, you'll need to dig up a user manual, to see what span
they can handle on mounting hardware. And since this is a Southbridge,
you also have to be careful that any cooler you buy, doesn't get
in the way of any video cards.

Oh, another thing.

http://www.supergluecorp.com/removingsuperglue.html

"even though Super Glue is incredibly strong, it has one weakness: acetone."

That is great news, except the plastic on the integrated circuits
might not be completely immune to acetone.

Paul

Paul nospam needed.com wrote:

.... see above

Hmmm. If you can't keep a finger on the Southbridge heatsink,
that means it is 65C+. Allowing for delta_T to the die, you
could be around 85 to 90C at the silicon die. Something is not
right there. Generally, if you burn yourself on something on
your motherboard, you need to "fix it".

Does your motherboard have any "boost voltage" settings for the
chipset ? It sounds like perhaps the core voltage on the
Southbridge is wrong. (In the manual, I see an "ICH I/O" and
"ICH Core" setting in the MB Intelligent Tweaker screen.)

The ICH10 datasheet has these values, so it takes a bunch of
different power inputs. I've marked the entries with arrows,
that I think correspond to the settings in the BIOS. So the
values shown in the manual (1.5V and 1.1V) could very well be
the correct stock settings.

"Intel ICH10 Absolute Maximum Ratings

1.1 V Supply Voltage with respect to VSS -0.5 to 2.1 V
--- 1.25 V Supply Voltage with respect to VSS -0.5V to
2.1V 1.5 V Supply Voltage with respect to VSS -0.5 to
2.1 V --- 3.3 V Supply Voltage with respect to VSS
-0.5 to 4.6 V 5.0 V Supply Voltage with respect to VSS
-0.5 to 5.5 V V_CPU_IO Supply Voltage with respect to VSS
-0.5 to 2.1 V"

"Vcc1_1 1.1 V supply for core well logic.
Vcc1_5_A 1.5 V supply for Logic and I/O."

Instead of Superglue, you should have used regular thermal
paste, and found another means of holding the heatsink to the
chip. The nice thing about thermal paste, is it doesn't need as
much "normal force" or "pressing downward force", to make the
grease a good conductor of heat. On chips which have bare die,
you have to be careful to use just enough grease, to allow the
silicon die to be covered when pressure is applied to the
heatsink. If the chip has a heat spreader lid on top, then you
can be a bit sloppier with the grease.

So you have two ways to attack the stability issue. Either
reduce the heat coming from the chip (turn down the core
voltage), or improve the cooling on the chip. I'm going to see
if I can find a picture of your motherboard, to see what kind of
heatsink is on it. If you use a heatsink with more cooling fin
area, then point a fan right at the fins (impingement cooling),
that is about the best you can do on the cooling front. You
might do even better with some kind of chilled liquid cooling
system, but for the price, buying another motherboard would be
cheaper.

OK, I can see the heatsink here.

http://images17.newegg.com/is/image/newegg/13-128-372-S03?$S640W$

You may be able to replace that with a Swiftech MCX159CU.

http://www.swiftnets.com/products/mcx159-CU.asp

"Brackets allow adjustments as follows:

center to center distance between mounting holes ranges
from 2.05" (52.1mm) to 2.41" (61.4mm)."

You measure the distance from the center of one pushpin to the
other, to see if the mounting scheme would work.

That isn't the absolute best one you can use. The problem with a
lot of those after-market ones, is they don't stay in production
forever.

OK, this is another style of cooler you could look at. This uses
a heatpipe, which means more of the heat makes it into the fins.
You can point a fan at one of these.

http://www.newegg.com/Product/Produc...82E16835233021

Thermalright makes some like that too. I think the Xigmatek is
a copy of the Thermalright. Thermalright probably made the first
ones of that style.

http://www.heatsinkfactory.com/therm...fx-chipset-coo
ler.html

With those, you'll need to dig up a user manual, to see what
span they can handle on mounting hardware. And since this is a
Southbridge, you also have to be careful that any cooler you
buy, doesn't get in the way of any video cards.

Oh, another thing.

http://www.supergluecorp.com/removingsuperglue.html

"even though Super Glue is incredibly strong, it has one
weakness: acetone."

That is great news, except the plastic on the integrated
circuits might not be completely immune to acetone.

Actually, the plastic on integrated circuits is probably not
immune to superglue. Have you ever noticed a tiny plume of smoke
rising from superglue? It is like an acid, it tends to melt
plastic. There is no removing the heatsink from my motherboard
now. I would not use superglue on plastic if it might need to be
undone sometime.

I have never seen anyone produce anything to conflict with the
idea that superglue is a good heat conductor. Have you ever
researched it? I get a grand total of two results for "superglue
conducts heat".

For "superglue doesn't conduct heat", one reply from an unknown is
"superglue doesn't conduct heat and it actually breaks down under
heat" but that is nonsense since some superglues can handle 250°F.
Another says "superglue doesn't conduct heat that well, even if it
did it would be in a thermal compound by now" but that is
obviously just ignorant. More accurate would be a statement like
"if superglue were determined to be a good heat sink compound,
heatsink compound producers would be out of business". One reason
superglue is not normally used is because, depending on the
surfaces, it is difficult or impossible to undo. And to the
unskilled, it can be very risky just trying to use it. Not just
the fact that it is so easy to get on your skin, but it is
something you have one shot at. Like skydiving... If at first you
don't succeed, skydiving is not for you.

I do not see anything scientific about the subject. I am surprised
some of the overclockers have not done a study or a simple
experiment.

Some superglues bond stuff with extremely little space in between
the two surfaces. Here are the specifications (if the following
link does not work, you can find the maximum separation distance
and other properties/information for various superglues at
McMaster Carr).

http://www.mcmaster.com/#loctite-super-glue/=9nx9j9

I guess they call it "fill space". Some superglue fills only
2/1000 of an inch, apparently the maximum. So if you have very
even surfaces, especially if they are slightly rough, the two
parts might even contact physically, especially after it dries and
pulls them tightly together.

And then there is the fact that you no longer need a clip or
anything to hold the heatsink to the chip.

Typical maximum temperature for superglue is probably around
180°F, according to their chart. That is only about 82°C. Good
thing the clips are still on it. But hey, maybe when the superglue
fails, it turns into a superconductor! Some of the Loctites on
that page go up to about 250°F.

I have some Arctic Silver, and the heatsink clips still worked,
but I wanted to experiment a bit (and put my money where my mouth
is, since I have mentioned using superglue before).

I am not sure that heat sinking was the problem, no errors in this
new installation so far, but I might as you suggest to try Messing
with the southbridge settings if they seem out of line.

John Doe wrote:
Paul nospam needed.com wrote:

... see above

Hmmm. If you can't keep a finger on the Southbridge heatsink,
that means it is 65C+. Allowing for delta_T to the die, you
could be around 85 to 90C at the silicon die. Something is not
right there. Generally, if you burn yourself on something on
your motherboard, you need to "fix it".

Does your motherboard have any "boost voltage" settings for the
chipset ? It sounds like perhaps the core voltage on the
Southbridge is wrong. (In the manual, I see an "ICH I/O" and
"ICH Core" setting in the MB Intelligent Tweaker screen.)

The ICH10 datasheet has these values, so it takes a bunch of
different power inputs. I've marked the entries with arrows,
that I think correspond to the settings in the BIOS. So the
values shown in the manual (1.5V and 1.1V) could very well be
the correct stock settings.

"Intel ICH10 Absolute Maximum Ratings

1.1 V Supply Voltage with respect to VSS -0.5 to 2.1 V
--- 1.25 V Supply Voltage with respect to VSS -0.5V to
2.1V 1.5 V Supply Voltage with respect to VSS -0.5 to
2.1 V --- 3.3 V Supply Voltage with respect to VSS
-0.5 to 4.6 V 5.0 V Supply Voltage with respect to VSS
-0.5 to 5.5 V V_CPU_IO Supply Voltage with respect to VSS
-0.5 to 2.1 V"

"Vcc1_1 1.1 V supply for core well logic.
Vcc1_5_A 1.5 V supply for Logic and I/O."

Instead of Superglue, you should have used regular thermal
paste, and found another means of holding the heatsink to the
chip. The nice thing about thermal paste, is it doesn't need as
much "normal force" or "pressing downward force", to make the
grease a good conductor of heat. On chips which have bare die,
you have to be careful to use just enough grease, to allow the
silicon die to be covered when pressure is applied to the
heatsink. If the chip has a heat spreader lid on top, then you
can be a bit sloppier with the grease.

So you have two ways to attack the stability issue. Either
reduce the heat coming from the chip (turn down the core
voltage), or improve the cooling on the chip. I'm going to see
if I can find a picture of your motherboard, to see what kind of
heatsink is on it. If you use a heatsink with more cooling fin
area, then point a fan right at the fins (impingement cooling),
that is about the best you can do on the cooling front. You
might do even better with some kind of chilled liquid cooling
system, but for the price, buying another motherboard would be
cheaper.

OK, I can see the heatsink here.

http://images17.newegg.com/is/image/newegg/13-128-372-S03?$S640W$

You may be able to replace that with a Swiftech MCX159CU.

http://www.swiftnets.com/products/mcx159-CU.asp

"Brackets allow adjustments as follows:

center to center distance between mounting holes ranges
from 2.05" (52.1mm) to 2.41" (61.4mm)."

You measure the distance from the center of one pushpin to the
other, to see if the mounting scheme would work.

That isn't the absolute best one you can use. The problem with a
lot of those after-market ones, is they don't stay in production
forever.

OK, this is another style of cooler you could look at. This uses
a heatpipe, which means more of the heat makes it into the fins.
You can point a fan at one of these.

http://www.newegg.com/Product/Produc...82E16835233021

Thermalright makes some like that too. I think the Xigmatek is
a copy of the Thermalright. Thermalright probably made the first
ones of that style.

http://www.heatsinkfactory.com/therm...fx-chipset-coo
ler.html

With those, you'll need to dig up a user manual, to see what
span they can handle on mounting hardware. And since this is a
Southbridge, you also have to be careful that any cooler you
buy, doesn't get in the way of any video cards.

Oh, another thing.

http://www.supergluecorp.com/removingsuperglue.html

"even though Super Glue is incredibly strong, it has one
weakness: acetone."

That is great news, except the plastic on the integrated
circuits might not be completely immune to acetone.

Actually, the plastic on integrated circuits is probably not
immune to superglue. Have you ever noticed a tiny plume of smoke
rising from superglue? It is like an acid, it tends to melt
plastic. There is no removing the heatsink from my motherboard
now. I would not use superglue on plastic if it might need to be
undone sometime.

I have never seen anyone produce anything to conflict with the
idea that superglue is a good heat conductor. Have you ever
researched it? I get a grand total of two results for "superglue
conducts heat".

For "superglue doesn't conduct heat", one reply from an unknown is
"superglue doesn't conduct heat and it actually breaks down under
heat" but that is nonsense since some superglues can handle 250°F.
Another says "superglue doesn't conduct heat that well, even if it
did it would be in a thermal compound by now" but that is
obviously just ignorant. More accurate would be a statement like
"if superglue were determined to be a good heat sink compound,
heatsink compound producers would be out of business". One reason
superglue is not normally used is because, depending on the
surfaces, it is difficult or impossible to undo. And to the
unskilled, it can be very risky just trying to use it. Not just
the fact that it is so easy to get on your skin, but it is
something you have one shot at. Like skydiving... If at first you
don't succeed, skydiving is not for you.

I do not see anything scientific about the subject. I am surprised
some of the overclockers have not done a study or a simple
experiment.

Some superglues bond stuff with extremely little space in between
the two surfaces. Here are the specifications (if the following
link does not work, you can find the maximum separation distance
and other properties/information for various superglues at
McMaster Carr).

http://www.mcmaster.com/#loctite-super-glue/=9nx9j9

I guess they call it "fill space". Some superglue fills only
2/1000 of an inch, apparently the maximum. So if you have very
even surfaces, especially if they are slightly rough, the two
parts might even contact physically, especially after it dries and
pulls them tightly together.

And then there is the fact that you no longer need a clip or
anything to hold the heatsink to the chip.

Typical maximum temperature for superglue is probably around
180°F, according to their chart. That is only about 82°C. Good
thing the clips are still on it. But hey, maybe when the superglue
fails, it turns into a superconductor! Some of the Loctites on
that page go up to about 250°F.

I have some Arctic Silver, and the heatsink clips still worked,
but I wanted to experiment a bit (and put my money where my mouth
is, since I have mentioned using superglue before).

I am not sure that heat sinking was the problem, no errors in this
new installation so far, but I might as you suggest to try Messing
with the southbridge settings if they seem out of line.

If you want any material to work as a heat conductor, you'd mix
boron nitride particles with it. That is what the thermal grease
compounds use as a major component. Even Arctic Silver has Boron
Nitride in it. Boron nitride is an abrasive, something you might
make sandpaper from. But it also conducts heat.

Instead of Superglue, you could use Thermal Epoxy. It is permanent.
One guy tried to separate the stuff, after it had set, and it
ripped the top off the IC. So the IC is weaker than the thermal
epoxy is. You only get one chance, to get it right, get the
heatsink on evenly and so on.

http://www.arcticsilver.com/arctic_s...l_adhesive.htm

The MSDS for that stuff, is here. So this is an epoxy kit, but
with Boron Nitride mixed into it. (I'm not sure what percentage
is silver particles.) The Boron Nitride has to have a small
particle size, to be effective.

http://www.arcticsilver.com/PDF/epoxy/ASTA_MSDS_3.pdf

"Product Ingredient Information CAS No.
Silver (Metallic) 7440-22-4
Boron Nitride 10043-11-5
Part A Resin (Liquid Plastic N.O.S.) Non Hazardous
Part B Hardener (Mercaptan Polymer) Non Hazardous"

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

"The thermal conductivity of BN is among the highest
of all electric insulators"

HTH,
Paul

Paul nospam needed.com wrote:

If you want any material to work as a heat conductor, you'd mix
boron nitride particles with it.

Closeness helps with heat conduction too.

Instead of Superglue,

Is my southbridge going to die?

Every review I have read that mentions using superglue says that
it works well. The only exception is trying to use the four corner
method with heat sink grease in the middle, and getting the two
mixed up. I would be concerned about that for conceivably
producing toxic fumes too.

John Doe wrote:
Paul nospam needed.com wrote:

If you want any material to work as a heat conductor, you'd mix
boron nitride particles with it.

Closeness helps with heat conduction too.

Instead of Superglue,

Is my southbridge going to die?

Every review I have read that mentions using superglue says that
it works well. The only exception is trying to use the four corner
method with heat sink grease in the middle, and getting the two
mixed up. I would be concerned about that for conceivably
producing toxic fumes too.

Have you worked on your BIOS settings yet ?

There is a difference between

ICH I/O [Auto]
ICH Core [Auto]

and

ICH I/O [1.500V]
ICH Core [1.100V]

Now, in theory, they're the same value. But
in practice, if there is a BIOS bug, the
Auto setting could actually be using a
radically different value for the voltage.
Forcing it manually, to the stock value,
might help.

I doubt the actual voltage value is measurable by
the hardware monitor chip, but check whatever
utility Gigabyte provides, to see if you can
monitor those voltages from Windows.

A Southbridge is good to about 99C at the silicon die. (You
can look up the exact value in the Intel documentation.) The
exterior of the heatsink clamped to the Southbridge, assuming
it's making good thermal contact, will be 20-25C cooler. I have
a digital thermometer I've used on occasion, that I jammed into
the fin array, to get some idea of the actual temperature. Then
do the math, add 20 to 25C to that number, and see if the result
is over 99C. Lots of laptops run the chipset that hot, for
their entire life, so they don't die instantly.

Paul

Paul nospam needed.com wrote:

Have you worked on your BIOS settings yet ?

Since reinstalling XP on my Raptor, there are no errors.

There is a difference between

ICH I/O [Auto]
ICH Core [Auto]

and

ICH I/O [1.500V]
ICH Core [1.100V]

Strangely, it goes from 1.45 to 1.55, but will not allow 1.5. I
will keep it on 1.45 for now. But I do not see the point unless
there are noticeable errors or unless the difference is noticeable
to the touch.

With backups of drive C, will keep saved copies of RootkitRevealer
results this time.

Now, in theory, they're the same value. But
in practice, if there is a BIOS bug, the
Auto setting could actually be using a
radically different value for the voltage.
Forcing it manually, to the stock value,
might help.

I doubt the actual voltage value is measurable by
the hardware monitor chip, but check whatever
utility Gigabyte provides, to see if you can
monitor those voltages from Windows.

Okey-dokey.
--















A Southbridge is good to about 99C at the silicon die. (You
can look up the exact value in the Intel documentation.) The
exterior of the heatsink clamped to the Southbridge, assuming
it's making good thermal contact, will be 20-25C cooler. I have
a digital thermometer I've used on occasion, that I jammed into
the fin array, to get some idea of the actual temperature. Then
do the math, add 20 to 25C to that number, and see if the result
is over 99C. Lots of laptops run the chipset that hot, for
their entire life, so they don't die instantly.

Paul