E4300 - Very fast temp change with loading

[Bob F]

In experimenting with overclocking a new system, I've noticed that
the CPU temperature displayed varies very quickly with load. Running
orthos gets the temp (displayed by PC Probe II) up to 129F within
a few minutes. Stopping orthos results in the temp dropping within
seconds to less than 100F. This suggests to me that there is poor
coupling between the heatsink and the processor, or internally within
the processor. I tried removing the origional heatsink "paste" from the
heatsink and processor and replacing it with some white dow corning
silicone HS paste, with no improvement.

It seems to me that the temp should vary much slower, as the heatsink
gets hotter or cooler. Can a different grade of HS paste make a big
difference in this, or am I interpreting the data wrong?

Bob F


Asus P5LD2
Intel E4300 - w/HS/fan from QX6700 Extreme Quad core
currently testing at 2.88 G OC
Crucial ValueSelect 2x1G DDR2-667
eVGA 7600GT Video
Maddog 16X DVD +-RW
WunXP Pro

 

[Phil Weldon]

'Bob F' wrote, in part:
| In experimenting with overclocking a new system, I've noticed that
| the CPU temperature displayed varies very quickly with load. Running
| orthos gets the temp (displayed by PC Probe II) up to 129F within
| a few minutes. Stopping orthos results in the temp dropping within
| seconds to less than 100F. This suggests to me that there is poor
| coupling between the heatsink and the processor, or internally within
| the processor.
_____

What you see is completely normal. The temperature measurement for Intel
CPUs comes from an on-CPU-die thermal diode. That temperature measurement
reacts immediately to increased current consumption on the CPU die. There
is a thermal resistance between the CPU die and the heat spreader; and an
additional thermal resistance between the heat spreader and the heat sink;
and an additional thermal resistance between the heatsink surface in contact
and the heat spreader; and an additional thermal resistance within the
heatsink; and additional thermal resistance between the heatsink and the
cooling fluid (air or water.)

The heat storage capacity of all of these elements before the cooling fluid
is very low - not nearly enough to slow the temperature rise significantly.
You may call that 'poor coupling' between the heatsink and the processor if
you wish, but that's the natural world. The temperatures you report are
pretty meaningless if you don't have the temperature of the cooling fluid
entering the heatsink. The interfaces between the CPU die and the heatsink
contact surface are only part of the total thermal resistance. If the CPU
is only issuing halt instructions, the power consumption is only a few
watts. In that case, the CPU temperature should only be a few degrees C
above the temperature of the cooling fluid entering the heatsink. The 43.5
C temperature you report at idle indicates the CPU is not issuing halt
instructions, but just running in an idle loop (unless the cooling fluid
temperature is ~ 40 C.) The 60 C temperature you report under load (if
'orthos' is a processor stress test) is probably normal for a plain vanilla
heatsink/fan and an Intel E4300 overclocked to 2.88 MHz.

The important factor in judging the cooling solution is the CPU temperature
compared with the cooling fluid temperature, NOT how quickly the CPU
temperature changes.

Thermal paste or thermal grease products all have very similar performance -
even butter works well as a thermal grease ( other than the fact it leaks
and goes rancid.) The highly hyped 'Arctic Silver' is only marginally
better than zinc oxide sunblock. More important than the brand you use is
how the thermal compound is applied, and how flat, smooth, and parallel the
surfaces are.

Phil Weldon



I tried removing the origional heatsink "paste" from the
| heatsink and processor and replacing it with some white dow corning
| silicone HS paste, with no improvement.
|
| It seems to me that the temp should vary much slower, as the heatsink
| gets hotter or cooler. Can a different grade of HS paste make a big
| difference in this, or am I interpreting the data wrong?
|
| Bob F
|
|
| Asus P5LD2
| Intel E4300 - w/HS/fan from QX6700 Extreme Quad core
| currently testing at 2.88 G OC
| Crucial ValueSelect 2x1G DDR2-667
| eVGA 7600GT Video
| Maddog 16X DVD +-RW
| WunXP Pro
|
|

 

[kony]

In experimenting with overclocking a new system, I've noticed that
the CPU temperature displayed varies very quickly with load. Running
orthos gets the temp (displayed by PC Probe II) up to 129F within
a few minutes. Stopping orthos results in the temp dropping within
seconds to less than 100F. This suggests to me that there is poor
coupling between the heatsink and the processor, or internally within
the processor. I tried removing the origional heatsink "paste" from the
heatsink and processor and replacing it with some white dow corning
silicone HS paste, with no improvement.

It seems to me that the temp should vary much slower, as the heatsink
gets hotter or cooler. Can a different grade of HS paste make a big
difference in this, or am I interpreting the data wrong?

Bob F


Asus P5LD2
Intel E4300 - w/HS/fan from QX6700 Extreme Quad core
currently testing at 2.88 G OC
Crucial ValueSelect 2x1G DDR2-667
eVGA 7600GT Video
Maddog 16X DVD +-RW
WunXP Pro

If you had an "ideal" interface where the CPU was somehow
embedded within the heatsink, you would be right (or closer
to that at least), but the die is very small and coupling to
the far larger heatsink is inherantly imperfect with a
properly installed heatsink, let alone one improperly
installed.

The better your HS paste is, the smaller the difference up
to a point, but you may already be at that point, or close
eough that further effort has too diminishing a return. The
key is not what the change is nor how fast, but whether the
CPU remains cool enough to be stable and have acceptible
lifespan when running at full load for an extended period of
time. 129F is about 54C (we usually note CPU temps in
Celcuis not Fahrenheit) which is cool enough.

 

[Ed Medlin]

In experimenting with overclocking a new system, I've noticed that
the CPU temperature displayed varies very quickly with load. Running
orthos gets the temp (displayed by PC Probe II) up to 129F within
a few minutes. Stopping orthos results in the temp dropping within
seconds to less than 100F. This suggests to me that there is poor
coupling between the heatsink and the processor, or internally within
the processor. I tried removing the origional heatsink "paste" from the
heatsink and processor and replacing it with some white dow corning
silicone HS paste, with no improvement.

It seems to me that the temp should vary much slower, as the heatsink
gets hotter or cooler. Can a different grade of HS paste make a big
difference in this, or am I interpreting the data wrong?

Bob F

Your results are perfectly normal with a stock hs/fan. You are well within
the operating temperature range of your processor. Please note that we
normally use Celsius (I kind of always just think in C when relating to
processor temps) instead of F. Your temps at the high end are low 50s C and
are just fine as long as everything is stable. You might be able to bring
the load temps down some with an aftermarket hs/fan if you want to spend a
few extra bucks. I see no need to do so unless you want to try and go a bit
higher with your OC. Another benefit is that many aftermarket soulutions are
quieter because of their use of larger and lower rpm fans. In my experience
Phil is absolutely correct as far as thermal compounds go. I normally use
the cheap stuff from Radio Shack. My local computer shop gives me a tube of
Artic Silver about everytime I buy anything there, so I use it because it is
free..... I see almost no difference between it and the cheap stuff.


Ed

 

[Bob F]

Thank you Phil, kony, and Ed. It would seem that what I am
seeing is probably mostly accounted for by the thermal resistance
within the chip, since the temp settles very quickly to what I guess
is the heatsink temp when unloading the processor.

Bob

 

[Richard Hopkins]

Thank you Phil, kony, and Ed. It would seem that what I am
seeing is probably mostly accounted for by the thermal
resistance within the chip,

You are probably also seeing the effects of the active processor multiplier
and voltage management ramping up the CPU's performance level.

While in an idle state, your processor should be running at reduced
clockspeed and core voltage. As soon as you load it up, it should pretty
much instantly revert to full speed/full voltage operation, with a
corresponding increase in heat output.
--


Richard Hopkins
Cardiff, Wales, United Kingdom
(replace nospam with pipex in reply address)