A hot running SL6PG

[Stephan Carydakis]

Hi All,

I have posted this question to another group
(alt.comp.hardware.overclocking) just recently. In an attempt to find
similar experiences or further suggestions, I thought I would post here as
well.

I recently purchased a used P4 SL6PG
(http://processorfinder.intel.com/Details.aspx?sSpec=SL6PG). Basically, the
idle temps are very high but the full load temps are within the Intel specs.
At first, with a stock cooler, I had socket temps of 58c idle / 64c full
load after running Prime95 for 2 or so hours and under-volted at 1.5 volts.
The mobo temps are 27/29, HDD are around 32 and the ambient temp is 23c. I
have 80mm inlet and exhaust fans. I purchased a Scythe Samurai Z in an
attempt to get lower temps and this was the case. The idle temp dropped to
51c and full load to 55c. Remember that these are socket sensor temps and
not CPU diode temps. Switching to the diode sensor adds 7 or 8 degrees
Celsius to the stated temps. Currently, the CPU diode temps read 58c idle
and 62c full load (again Prime95 maximum heat tests for over 3 hours for the
load temps).

My system seems to be very stable however I am concerned with the high idle
temp. Even though I have a stable system, does anyone's experienced opinion
say that these readings indicate a 'badly' overclocked (insufficient
cooling, over volted etc) CPU? I'm hesitant to believe that the mobo is
giving false or inaccurate readings because the CPU that came out of the
same system, a P4 2.4b @ 2.88 and 1.6 volts, ran at 49c idle / 64c full load
on stock Intel cooling.

My OS is Win2k AS and I use Motherboard Monitor 5 to get these readings.

Thanks in advance,

Steph.

 

[paulmd]

Hi All,

I have posted this question to another group
(alt.comp.hardware.overclocking) just recently. In an attempt to find
similar experiences or further suggestions, I thought I would post here as
well.

I recently purchased a used P4 SL6PG
(http://processorfinder.intel.com/Details.aspx?sSpec=SL6PG). Basically, the
idle temps are very high but the full load temps are within the Intel specs.
At first, with a stock cooler, I had socket temps of 58c idle / 64c full
load after running Prime95 for 2 or so hours and under-volted at 1.5 volts.
The mobo temps are 27/29, HDD are around 32 and the ambient temp is 23c. I
have 80mm inlet and exhaust fans. I purchased a Scythe Samurai Z in an
attempt to get lower temps and this was the case. The idle temp dropped to
51c and full load to 55c. Remember that these are socket sensor temps and
not CPU diode temps. Switching to the diode sensor adds 7 or 8 degrees
Celsius to the stated temps. Currently, the CPU diode temps read 58c idle
and 62c full load (again Prime95 maximum heat tests for over 3 hours for the
load temps).

My system seems to be very stable however I am concerned with the high idle
temp. Even though I have a stable system, does anyone's experienced opinion
say that these readings indicate a 'badly' overclocked (insufficient
cooling, over volted etc) CPU? I'm hesitant to believe that the mobo is
giving false or inaccurate readings because the CPU that came out of the
same system, a P4 2.4b @ 2.88 and 1.6 volts, ran at 49c idle / 64c full load
on stock Intel cooling.

My OS is Win2k AS and I use Motherboard Monitor 5 to get these readings.

Thanks in advance,

Steph.-

Got dust? If so blow it out.

If not, then it's a question of redoing the heatsink/chip interface by
cleaning off the old thermal compound, and reapplying arctic silver,
getting a better heatsink, or case fans, a new case with better
airflow, or any combination of the above.

The chip will sive longer if it runs cooler, even if you ARE getting a
false reading, cooling it more would be good.

PS 64c at full load is a LOT on the toasty side, too.

 

[Rod Speed]

(e-mail address removed) wrote

Stephan Carydakis wrote

Got dust? If so blow it out.

Unlikely that dust would affect just the idle temp and not the full load temp.

 

[kony]

Hi All,

I have posted this question to another group
(alt.comp.hardware.overclocking) just recently. In an attempt to find
similar experiences or further suggestions, I thought I would post here as
well.

I recently purchased a used P4 SL6PG
(http://processorfinder.intel.com/Details.aspx?sSpec=SL6PG). Basically, the
idle temps are very high but the full load temps are within the Intel specs.
At first, with a stock cooler, I had socket temps of 58c idle / 64c full
load after running Prime95 for 2 or so hours and under-volted at 1.5 volts.
The mobo temps are 27/29, HDD are around 32 and the ambient temp is 23c. I
have 80mm inlet and exhaust fans. I purchased a Scythe Samurai Z in an
attempt to get lower temps and this was the case. The idle temp dropped to
51c and full load to 55c. Remember that these are socket sensor temps and
not CPU diode temps. Switching to the diode sensor adds 7 or 8 degrees
Celsius to the stated temps. Currently, the CPU diode temps read 58c idle
and 62c full load (again Prime95 maximum heat tests for over 3 hours for the
load temps).

My system seems to be very stable however I am concerned with the high idle
temp. Even though I have a stable system, does anyone's experienced opinion
say that these readings indicate a 'badly' overclocked (insufficient
cooling, over volted etc) CPU?

No, an overclocked or insufficiently cooled CPU will still
have a similarly large difference between idle and full
load.

It is more likely either:

A) Something in the background is keeping CPU busy, perhaps
a badly written app, a monitoring program, spyware/viri/etc,
or ???

B) The system wasn't installed with ACPI functional so
there is no HALT-cooling, happening.

I'm hesitant to believe that the mobo is
giving false or inaccurate readings because the CPU that came out of the
same system, a P4 2.4b @ 2.88 and 1.6 volts, ran at 49c idle / 64c full load
on stock Intel cooling.

First, what are you using to generate what you call the
"full load"? If it isn't really a full load, that migth
explain the small difference, but even so, 49/64 isn't such
a small difference.

My OS is Win2k AS and I use Motherboard Monitor 5 to get these readings.

I suggest using something other than MB5, to get an accurate
reading you might need a temp offset of some sort which it
may not provide. Check against the BIOS temp, it should be
(rough guesstimation) a little closer to, but not quite, the
full load temp, rather than the idle-with-ACPI_OS temp.

 

[Stephan Carydakis]

Hi,

Thanks for your time.

As mentioned, I'm using the Prime95 torture test with the maximum heat
option and run it for 2 or 3 hours to get a full load test. I also
purchased a new Scythe Samurai Z and this did bring the temps down 4c on
idle and more on full load. There is no dust as this is now a brand new
cooler. Previous to the new cooler, I did regularly cleaned the dust from
the stock cooler. With the new CPU, I did try reinstalling the stock cooler
3 times using Arctic Silver 5. This did not make a difference to temps.

Again, the previous CPU that, a P4 2.4b @ 2.88 and 1.6 volts, ran a lot
cooler at idle on stock cooling so I'm using this as a benchmark for the new
temps (even though the thermal output of the new CPU will be more). Because
my mobo and hdd temps are still as they were before the new CPU was
installed, I'm reluctant to believe the problem is false readings although
anything is possible.

Your suggestions make good sense but because the old CPU ran a lot cooler
OC'd, with a higher voltage, and on a stock cooler, as well as the mobo/hdd
temps being the same as previous readings, I'm hesitant to believe that my
case or case airflow is an issue. With front and rear 80mm fans and mobo
temps of mid to high 20's, this tells me the air flow is probably good!

Thanks again,

Steph.

 

[Stephan Carydakis]

Hi,

Thanks for you response.

No, an overclocked or insufficiently cooled CPU will still
have a similarly large difference between idle and full
load.

What I meant is that if this CPU was previosly mistreated, as in oc'd with
insufficient cooling or generally oc'd improperly, could this have damage to
the CPU, and caused it to have such a small spread in temps?

It is more likely either:

A) Something in the background is keeping CPU busy, perhaps
a badly written app, a monitoring program, spyware/viri/etc,
or ???

B) The system wasn't installed with ACPI functional so
there is no HALT-cooling, happening.

But it is now!

First, what are you using to generate what you call the
"full load"? If it isn't really a full load, that migth
explain the small difference, but even so, 49/64 isn't such
a small difference.

Prime95 torture test with the maximum heat option for more than 2 hours. The
49c was my previous CPU idle temp (on chip sensor) not from the CPU in
question. Right now, my new CPU is OC'd to 3450 MHz and 1.55v and is 61c at
idle and does not go beyond 65c after 3 hours of Prime95 max heat torture
test. At stock speed, its around 58c idle 62c full load. These are from the
on chip sensor, not the socket sensor on the mobo. The socket sensor
currently reads 52c (9c lower than the on chip sensor) which coincides with
the BIOS readings. This is still very high.

I suggest using something other than MB5, to get an accurate
reading you might need a temp offset of some sort which it
may not provide. Check against the BIOS temp, it should be
(rough guesstimation) a little closer to, but not quite, the
full load temp, rather than the idle-with-ACPI_OS temp.

Everyone I have spoken to says to use MB5? Can you suggest something that
may be more accurate?

Thanks again,

Steph.

 

[Rod Speed]

Thanks for your time.

As mentioned, I'm using the Prime95 torture test with the maximum
heat option and run it for 2 or 3 hours to get a full load test. I also
purchased a new Scythe Samurai Z and this did bring the temps
down 4c on idle and more on full load. There is no dust as this is
now a brand new cooler. Previous to the new cooler, I did regularly
cleaned the dust from the stock cooler. With the new CPU, I did try
reinstalling the stock cooler 3 times using Arctic Silver 5. This did
not make a difference to temps.

Again, the previous CPU that, a P4 2.4b @ 2.88 and 1.6 volts, ran a
lot cooler at idle on stock cooling so I'm using this as a benchmark
for the new temps (even though the thermal output of the new CPU will
be more). Because my mobo and hdd temps are still as they were before
the new CPU was installed, I'm reluctant to believe the problem is
false readings although anything is possible.

Your suggestions make good sense but because the old CPU ran a lot
cooler OC'd, with a higher voltage, and on a stock cooler, as well as
the mobo/hdd temps being the same as previous readings,

Most likely the real difference between the two idle temps
is that the original is a Northwood cpu and the latests is a
Prescott which are rather notorious for running much warmer.

I'm hesitant to believe that my case or case airflow is an issue.

It can be with Prescott cpus. They work best when configured so that
external air is used directly onto the cpu heatsink and fan. What is
viable with a Northwood wont necessarily work that well with a Prescott.

With front and rear 80mm fans and mobo temps of mid
to high 20's, this tells me the air flow is probably good!

Looks like it isnt with a Prescott cpu.

 

[kony]

What I meant is that if this CPU was previosly mistreated, as in oc'd with
insufficient cooling or generally oc'd improperly, could this have damage to
the CPU, and caused it to have such a small spread in temps?

MIstreatment and/or overclocking could damage a CPU (depends
a LOT on exactly what had happened) but this is not
expected, nor can I think of any situation where it would
result in the small spread in temps.

In general, no matter what the prior treatment of it was, if
it is now computationally intact, there should be nothing
that would cause it. Perhaps long term overheating would
damage the core or carrier to the extent that eventually the
CPU would fail, but until it did fail, it would work
normally. We don't necessarily know if it's working
normally though, extensive stress testing including results
checking (like Prime95 does) would be one applicable test.

But it is now!

It is now what?
If the OS was installed while ACPI was disabled, merely
enabling it in the bios won't be enough, the OS would need
reinstalled. There may be a way to hack the OS to enable it
later but that is experimental at best.

Prime95 torture test with the maximum heat option for more than 2 hours. The
49c was my previous CPU idle temp (on chip sensor) not from the CPU in
question. Right now, my new CPU is OC'd to 3450 MHz and 1.55v and is 61c at
idle and does not go beyond 65c after 3 hours of Prime95 max heat torture
test. At stock speed, its around 58c idle 62c full load. These are from the
on chip sensor, not the socket sensor on the mobo. The socket sensor
currently reads 52c (9c lower than the on chip sensor) which coincides with
the BIOS readings. This is still very high.

I'm wondering if you're not reading the CPU temp at all.
"IF" ACPI is enabled and working properly, AND there's
nothing extra running in the background to load the CPU
(Task manager shows a few low, single-digit % utilization),
and your full load app works properly (as I presume it does,
and as Task Manager can confirm enough for this purpose),
then to have only 4C temp change suggests you are not
monitoring the CPU temp, that motherboard monitor is giving
the reading of a different chip, perhaps the northbridge.

This is why it is good to use a motherboard specific
utility, but even then, occasionally the utility is older
than the board and neeed patched (just newer version of the
whole utility, usually) to report correctly.

Everyone I have spoken to says to use MB5? Can you suggest something that
may be more accurate?

Do as mentioned previously, compare to the bios reported
temp.

MBM5 may work but you will have to fiddle with the settings
to find all temp reports and compare how those reported
temps change with changes in load.

 

[paulmd]

Everyone I have spoken to says to use MB5? Can you suggest something that
may be more accurate?

Thanks again,

Steph.

We can argue about the exact cause, or we can just say "chip too hot,
need better cooling <grunt>"

In any event, the temperature has got to come down.

One final question: is your power supply adequite?

 

[Sjouke Burry]

MB5 reports:case 12C CPU 36C sensor 30C
with my room temp at 25C
I don't think thats a usable diagnostic.

Speedfan 4.27:
temp2 46C
temp3 37C
HD0 39C
HD1 28C
Now that looks better,HD1 is in sleep
mode,so only 3C above ambient.
It has also a nice SMART readout of the
disks,to let you check disk health.

 

[Stephan Carydakis]

MIstreatment and/or overclocking could damage a CPU (depends
a LOT on exactly what had happened) but this is not
expected, nor can I think of any situation where it would
result in the small spread in temps.

In general, no matter what the prior treatment of it was, if
it is now computationally intact, there should be nothing
that would cause it. Perhaps long term overheating would
damage the core or carrier to the extent that eventually the
CPU would fail, but until it did fail, it would work
normally. We don't necessarily know if it's working
normally though, extensive stress testing including results
checking (like Prime95 does) would be one applicable test.



It is now what?
If the OS was installed while ACPI was disabled, merely
enabling it in the bios won't be enough, the OS would need
reinstalled. There may be a way to hack the OS to enable it
later but that is experimental at best.



I'm wondering if you're not reading the CPU temp at all.
"IF" ACPI is enabled and working properly, AND there's
nothing extra running in the background to load the CPU
(Task manager shows a few low, single-digit % utilization),
and your full load app works properly (as I presume it does,
and as Task Manager can confirm enough for this purpose),
then to have only 4C temp change suggests you are not
monitoring the CPU temp, that motherboard monitor is giving
the reading of a different chip, perhaps the northbridge.

This is why it is good to use a motherboard specific
utility, but even then, occasionally the utility is older
than the board and neeed patched (just newer version of the
whole utility, usually) to report correctly.

These are the same sensors I used for my previous P4 2.4b and the temp
spread on that was more. More like 49c/65c (oc'd 2.88 @ 1.6V) from the diode
sensor. As I mentioned, using the diode sensor gives me about 8c or 9c
higher readings than the socket sensor. The full load app is Prime95 using
the maximum heat test for 2 or more hours.

Do as mentioned previously, compare to the bios reported
temp.

The bios temp seems to coincide with the socket sensor temps MB5 is giving
me.

 

[Stephan Carydakis]

We can argue about the exact cause, or we can just say "chip too hot,
need better cooling <grunt>"

In any event, the temperature has got to come down.

One final question: is your power supply adequite?

Maybe some pvc piping directing air from my 3hp split system (running in
cooling mode of course) will make all my computer troubles go away. Of
course, it costs about $7 a day to run and would send me broke. Interesting
what you ask about my PSU. No one had suggested that my PSU may be the
culprit. It was with my previous CPU and seems to be with this one but how
do I determine that? It's nothing special - a Juster 400w unit.

 

[paulmd]

Maybe some pvc piping directing air from my 3hp split system (running in
cooling mode of course) will make all my computer troubles go away. Of
course, it costs about $7 a day to run and would send me broke. Interesting
what you ask about my PSU. No one had suggested that my PSU may be the
culprit. It was with my previous CPU and seems to be with this one but how
do I determine that? It's nothing special - a Juster 400w unit.

Probly better for a single one-time expense of an oversized heatsink.

400W is probably OK unless you're running a large number of drives.

Though if the fan(s) on your power supply died, it could be a major
contributer to an airflow problem.

 

[kony]

Maybe some pvc piping directing air from my 3hp split system (running in
cooling mode of course) will make all my computer troubles go away. Of
course, it costs about $7 a day to run and would send me broke. Interesting
what you ask about my PSU. No one had suggested that my PSU may be the
culprit. It was with my previous CPU and seems to be with this one but how
do I determine that? It's nothing special - a Juster 400w unit.

I don't know what a "Juster" is, though likely generic and
questionable. Even so, unless the fan had failed as Paul
mentioned, it should not account fo r higher temps nor a
small spread between idle and load.

Consider that many people run a similar CPU without
overheating. Scrutinize each system, cooling subsystem to
see which one (if not multiple) is the culprit. While some
CPU like a P4 Prescott are hot running, they're still
manageable without extraordinary measures.

 

[Stephan Carydakis]

I don't know what a "Juster" is, though likely generic and
questionable. Even so, unless the fan had failed as Paul
mentioned, it should not account fo r higher temps nor a
small spread between idle and load.

Consider that many people run a similar CPU without
overheating. Scrutinize each system, cooling subsystem to
see which one (if not multiple) is the culprit. While some
CPU like a P4 Prescott are hot running, they're still
manageable without extraordinary measures.

I think have reasonably looked at most possibilities. So far I have:

1. Re installed the CPU on stock cooling 3 times: no difference
2. Purchased and installed a Scythe Samurai Z CPU cooler using Arctic
Silver 5: 4c drop in idle and max temps
3. Used a different thermal compound on new cooler: no difference
4. Undervolted and under clocked (1.45v and 2.53GHz): 1-2c drop in idle and
max temps
5. Checked PSU, front and rear case fans. All functioning.

Considering the internal case temps are not high and the mobo temp reading
is 25c with an ambient of 21c, I dont think airflow is a problem. Other
components like the HDD also run very cool at 30c. With the rear fan at
5000rpm (and a whole heap of noise) I can get the case temp to 1c above
ambient and the CPU idle/max drops 2c after 5 or so minutes. After 15
minutes, the mobo temp is reading the same as my ambient temp. HDDs are at
28 and the CPU is at 59c (OC'd 3496 and a stock 1.55v).

I have to think this is a CPU problem. The only 2 things I can see that may
make a difference is swapping the PSU or the CPU. Since these CPUs are still
saleable, I think I will get another one. If this runs cooler, it was the
CPU! If not, I can sell it back on eBay and I may consider a new PSU.

Thanks to everyone for their suggestions and input.

 

[JohnH]

I think have reasonably looked at most possibilities. So far I have:

1. Re installed the CPU on stock cooling 3 times: no difference
2. Purchased and installed a Scythe Samurai Z CPU cooler using Arctic
Silver 5: 4c drop in idle and max temps
3. Used a different thermal compound on new cooler: no difference
4. Undervolted and under clocked (1.45v and 2.53GHz): 1-2c drop in
idle and max temps
5. Checked PSU, front and rear case fans. All functioning.

Considering the internal case temps are not high and the mobo temp
reading is 25c with an ambient of 21c, I dont think airflow is a problem.

Bet it is. Prescotts get pretty warm and work best with
ambient air into the heatsink and fan, not case air.

Other components like the HDD also run very cool at 30c.
With the rear fan at 5000rpm (and a whole heap of noise) I can get
the case temp to 1c above ambient and the CPU idle/max drops 2c after
5 or so minutes. After 15 minutes, the mobo temp is reading the same
as my ambient temp. HDDs are at 28 and the CPU is at 59c (OC'd 3496
and a stock 1.55v).

I have to think this is a CPU problem. The only 2 things I can see
that may make a difference is swapping the PSU or the CPU.

Try running with the case covers off.

Since these CPUs are still saleable, I think I will get another one.
If this runs cooler, it was the CPU! If not, I can sell it back on eBay

Unlikely to be the cpu. Plenty of Prescotts without external air
onto the cpu heatsink and fan get as hot as yours does at idle.

and I may consider a new PSU.

It wont be that either.

 

[Stephan Carydakis]

Oops. Reply to sender isn't the same as reply to group. Missed it by that
much...sorry about that Chief.

----- Original Message -----
From: "JohnH" <[email protected]>
Newsgroups: alt.comp.hardware
Sent: Tuesday, August 01, 2006 4:33 PM
Subject: Re: A hot running SL6PG

problem.

Bet it is. Prescotts get pretty warm and work best with
ambient air into the heatsink and fan, not case air.

Bet it isn't! It IS NOT a Prescott! Its a Northwood. See attachment on
previous post. An SL6PG is a Northwood.

Try running with the case covers off.

I have run it with covers off a few times. There is a small drop in temps,
similar results to running the rear case fan at 3000+rpm. But nothing huge.

Unlikely to be the cpu. Plenty of Prescotts without external air
onto the cpu heatsink and fan get as hot as yours does at idle.

That maybe true for a Prescott. The Thermal Spec on this CPU is 81.8w - not
THAT hot.

It wont be that either.

In my relatively inexperienced opinion, I wouldn't of thought so either but
I have read reports of people swapping out their PSU for a quality unit and
having their overheating problems go away? Don't know about the accuracy of
such reports.

 

[kony]

5. Checked PSU, front and rear case fans. All functioning.

It is not only a matter of them spinning but the rate at
which they're actually moving air as well as their vicininty
to the hot parts.

Suppose your front or rear fans have grills with
significant impedance to the airflow... that'll make the
fans far less effective.

In most cases, a good way to keep the Prescott temps down is
to make sure the case's rear exhaust fan reduces the air
recycled through the CPU 'sink (having gone through the
heatsink and been heated up some already, it should be
exhausted out the rear of the chassis as much as possible
instead of sucked back into the 'sink.

Considering the internal case temps are not high and the mobo temp reading
is 25c with an ambient of 21c, I dont think airflow is a problem. Other
components like the HDD also run very cool at 30c. With the rear fan at
5000rpm (and a whole heap of noise) I can get the case temp to 1c above
ambient and the CPU idle/max drops 2c after 5 or so minutes. After 15
minutes, the mobo temp is reading the same as my ambient temp. HDDs are at
28 and the CPU is at 59c (OC'd 3496 and a stock 1.55v).

If your chassis ambient temp is actually 25C, your heatsink
is not doing very well. I don't have that 'sink but I dont'
think much of any other scythe products so it's a big
variable...

I have to think this is a CPU problem.

Very unlikely.
If the CPU were shorting out the whole thing would be
failing, it wouldn't just run a little bit hotter.

If you have a multimeter you could try measuring the CPU
vcore voltage coming from the motherboard step-down supply
to see if it's at the correct voltage as you'd set it. Some
drift a bit higher, and if high enough it might make a
difference but since you were undervolting already, I'd
still most likely suspect the heatsink or chassis cooling,
not the CPU unless there is some surface defect on the heat
spreader interfering with close(r) contact to the 'sink.

The only 2 things I can see that may
make a difference is swapping the PSU or the CPU. Since these CPUs are still
saleable, I think I will get another one. If this runs cooler, it was the
CPU! If not, I can sell it back on eBay and I may consider a new PSU.

Neither should make a difference.
Your problem is the heat is not being removed efficiently
enough, quickly enough. Maybe the heat spreader isn't on
good... I've heard of people turning on a system with a p4
but the heatsink wasn't on, and the CPU heated up so quick
it popped the core away from the spreader, leaving a small
air-gap. That gap would be compressed a lot when the
heatsink is later mounted, "maybe" allowing it to run ok but
since your temps are only a dozen degrees above expected
levels I don't think that has happened.

 

[JohnH]

Oops. Reply to sender isn't the same as reply to group. Missed it by
that much...sorry about that Chief.

----- Original Message -----
From: "JohnH" <[email protected]>
Newsgroups: alt.comp.hardware
Sent: Tuesday, August 01, 2006 4:33 PM
Subject: Re: A hot running SL6PG



Bet it isn't! It IS NOT a Prescott! Its a Northwood. See attachment
on previous post. An SL6PG is a Northwood.


I have run it with covers off a few times. There is a small drop in
temps, similar results to running the rear case fan at 3000+rpm. But
nothing huge.


That maybe true for a Prescott. The Thermal Spec on this CPU is 81.8w
- not THAT hot.


In my relatively inexperienced opinion, I wouldn't of thought so
either but I have read reports of people swapping out their PSU
for a quality unit and having their overheating problems go away?

Don't believe it myself. There's a fundamental problem with the basic physics.

Don't know about the accuracy of such reports.

Pure fantasy, basically.

 

[Stephan Carydakis]

It is not only a matter of them spinning but the rate at
which they're actually moving air as well as their vicininty
to the hot parts.

Suppose your front or rear fans have grills with
significant impedance to the airflow... that'll make the
fans far less effective.

In most cases, a good way to keep the Prescott temps down is
to make sure the case's rear exhaust fan reduces the air
recycled through the CPU 'sink (having gone through the
heatsink and been heated up some already, it should be
exhausted out the rear of the chassis as much as possible
instead of sucked back into the 'sink.

I realise that a fan design will affect cfm but in general, push up the rpm
and the cfm goes up with it yes? I think the evidence points to good air
flow. I just cranked the RPM on the exhaust fan to 4500 (usually run it at
2000) and in 4 minutes and 47 seconds, the mobo/case temp settled at the
same as the room temp (from 25 to 21). The CPU and HDD temps also dropped by
2c. There is a filter on the front fan which cools the HDDs but this is
cleaned regularly. There is plenty of other holes for the exhaust fan to
suck in cooler room air anyway. Also, the exhaust fan is only centimeters
away from the CPU heatsink.

If your chassis ambient temp is actually 25C, your heatsink
is not doing very well. I don't have that 'sink but I dont'
think much of any other scythe products so it's a big
variable...

Why is that? Which heatsink? The Northbridge? I think 4c above the outside
air temp for a mobo reading is good actually. The Scythe Samaurai Z would be
what I consider only a couple of points up from stock cooling. As I
mentioned, it did bring the idle temp down 4c and the full load 7c. No bad
for such a quite fan. I'm sure with a larger 120mm or faster spinning/more
efficient 92mm fan it would work even better.

Very unlikely.
If the CPU were shorting out the whole thing would be
failing, it wouldn't just run a little bit hotter.

And that it is!

If you have a multimeter you could try measuring the CPU
vcore voltage coming from the motherboard step-down supply
to see if it's at the correct voltage as you'd set it. Some
drift a bit higher, and if high enough it might make a
difference but since you were undervolting already, I'd
still most likely suspect the heatsink or chassis cooling,
not the CPU unless there is some surface defect on the heat
spreader interfering with close(r) contact to the 'sink.


Neither should make a difference.
Your problem is the heat is not being removed efficiently
enough, quickly enough. Maybe the heat spreader isn't on
good... I've heard of people turning on a system with a p4
but the heatsink wasn't on, and the CPU heated up so quick
it popped the core away from the spreader, leaving a small
air-gap. That gap would be compressed a lot when the
heatsink is later mounted, "maybe" allowing it to run ok but
since your temps are only a dozen degrees above expected
levels I don't think that has happened.

I really can't see that a difference of 28 watts (between the old and new
CPU) would cause so many more c's at idle, especially considering I didn't
have an exhaust fan with the old CPU as well as having an aftermarket cooler
with the new CPU which has proven to be better than stock. Doesn't make
sense. But I am not an expert in PC cooling. Anyways, If the new CPU (given
I can get one at a reasonable price) behaves the same, I will concede that I
have a cooling problem! Thanks for the food for thought.