Intel desktop utilities and thermal margin for E8400 cpu

[Ken]

I just built a new system with an E8400 cpu on an Intel DP35DP board.
When I installed Desktop Utilities, it displays the processor
temperature as a thermal margin, in this case, 60 degrees centigrade.

I assume this is a pretty good margin of safety, IF the measurement is
correct. How does this translate into the actual temperature of the
cpu? This seems too good for just the stock Intel heatsink and fan from
what I have heard. Is Desktop Utilities accurate?

Thanks
Ken K

 

[Paul]

I just built a new system with an E8400 cpu on an Intel DP35DP board.
When I installed Desktop Utilities, it displays the processor
temperature as a thermal margin, in this case, 60 degrees centigrade.

I assume this is a pretty good margin of safety, IF the measurement is
correct. How does this translate into the actual temperature of the
cpu? This seems too good for just the stock Intel heatsink and fan from
what I have heard. Is Desktop Utilities accurate?

Thanks
Ken K

They state it as a "thermal margin" for a reason. It is because
the Core Temp digital temperature circuit, measures with respect
to Tcase_max. It is not an absolute temperature measurement as
such. It is a relative measurement.

In other words, the readout on the thing says "we are currently 60C
below the point at which I would throttle". You need to look up
Tcase_max, to compute an approximate absolute temperature. Say
Tcase_max was 100C. Then 100C - 60C = 40C absolute temperature.
If the "thermal margin" is reduced to zero, the thing is at
100C, and at that point, the CPU begins to use throttling to
try to reduce the temperature. Throttling reduces your
computing performance.

I read a posting from someone, who has attempted to calibrate the
Intel sensor. His finding is, the thing is most accurate at high
temperatures. And the error in the measurement is quite far off,
when the CPU is not in any thermal trouble. (I.e. Could easily
be off by 10C, when the CPU is quite cool.) So the thing is not
really an improvement over any previous temperature measurement
methods. Just because it is digital, doesn't mean it is
wonderful. It still has analog underpinnings at some level in
the silicon design.

The thermal diode, connected to the hardware monitor interface
on the motherboard, will still function as a second measurement
method.

Also, a problem with new processor introductions, is Tcase_max
is not the same across the entire product line. So when using
this relative measurement method, you're at the mercy of
Intel to give you the actual value they use. As it is, people
are currently guessing at some of them. Hardly an ideal
situation. Intel should just post a table of values, telling
you that the Tcase_max for "E8400 = 100C" or whatever.

Paul

 

[Ken]

Paul said the following on 4/27/2008 1:32 PM:

They state it as a "thermal margin" for a reason. It is because
the Core Temp digital temperature circuit, measures with respect
to Tcase_max. It is not an absolute temperature measurement as
such. It is a relative measurement.

In other words, the readout on the thing says "we are currently 60C
below the point at which I would throttle". You need to look up
Tcase_max, to compute an approximate absolute temperature. Say
Tcase_max was 100C. Then 100C - 60C = 40C absolute temperature.
If the "thermal margin" is reduced to zero, the thing is at
100C, and at that point, the CPU begins to use throttling to
try to reduce the temperature. Throttling reduces your
computing performance.

I read a posting from someone, who has attempted to calibrate the
Intel sensor. His finding is, the thing is most accurate at high
temperatures. And the error in the measurement is quite far off,
when the CPU is not in any thermal trouble. (I.e. Could easily
be off by 10C, when the CPU is quite cool.) So the thing is not
really an improvement over any previous temperature measurement
methods. Just because it is digital, doesn't mean it is
wonderful. It still has analog underpinnings at some level in
the silicon design.

The thermal diode, connected to the hardware monitor interface
on the motherboard, will still function as a second measurement
method.

Also, a problem with new processor introductions, is Tcase_max
is not the same across the entire product line. So when using
this relative measurement method, you're at the mercy of
Intel to give you the actual value they use. As it is, people
are currently guessing at some of them. Hardly an ideal
situation. Intel should just post a table of values, telling
you that the Tcase_max for "E8400 = 100C" or whatever.

Paul

I have a digital thermometer with a flexible probe (I used it for
measuring the temps at various parts of my espresso machine to be sure
the temp of the water was 203F. I will see if I can get to the heatsink
without shorting out the mb to see what the temp of the heatsink is...