*LOW* Athlon XP temperature!?!

[dave_bonnell]

I found something "interesting" last night, as I built up an Athlon XP
1800+ system. The processor was previously installed on an ASUS board.
Idle temperature was about 55 degrees. I then put some Arctic Silver
5 on, and temps dropped to 49-50 degrees. Note that this reading was
taken from the BIOS....the AsusProbe software consistently reported
about 40 degrees (I think this is socket temperature, or perhaps
something else).

So last night I put the CPU in a MSI nForce2 board. I installed the
same stock HSF. I started the system and proceeded to format a hard
drive, which took about 20 minutes. Then, I rebooted the system,
entered the BIOS, and checked the temperature:

System: 27 degrees
CPU: 19 degrees

WTF?!?! Something's not right here. I do a double-take. At first,
I'm suspicious of the board, which I purchased second-hand. My system
temperature using the ASUS board was only 23 degrees. Clearly the CPU
temperature is not correct....or is it?

I "check" the temperature by grasping the heatsink with my fingers. It
is cool to the touch! This is odd, since I'm sure it was warm in the
ASUS system. As a sanity check, I unplug the CPU fan (which, according
to the BIOS is running at 4000 RPM) and watch the temperature.
S-l-o-w-l-y, the temp rises, and I continue to monitor the HSF with my
fingers. By the time the temp reads 30 degrees, the HSF is warm enough
to believe that a processor is actually running (and that the
temperature reading is indeed incorrect).

Somewhat confused, I plugged the CPU fan back in and proceeded to
install Windows. Once installed, I let the system run on idle for a
while....and the heatsink is still cool to the touch!

Has anyone else experienced this sort of thing? I upgraded my ASUS rig
to a Bartion 2600, overclocked to 2075 MHz (instead of the normal 1909
MHz). The BIOS indicates a 55 degree temperature, or approximately 5
degree increase. I think I will try this processor in the MSI system
next, to get an idea of how the temp sensor operates.

Cheers,
Dave

 

[Jamie]

Personally I only take temps with a grain of salt I don't believe any are
very accurate. Anybody know of a sensor or and board or board utility that
is accurate?

 

[John]

I found something "interesting" last night, as I built up an Athlon XP
1800+ system. The processor was previously installed on an ASUS board.
Idle temperature was about 55 degrees. I then put some Arctic Silver
5 on, and temps dropped to 49-50 degrees. Note that this reading was
taken from the BIOS....the AsusProbe software consistently reported
about 40 degrees (I think this is socket temperature, or perhaps
something else).

Has anyone else experienced this sort of thing? I upgraded my ASUS rig
to a Bartion 2600, overclocked to 2075 MHz (instead of the normal 1909
MHz). The BIOS indicates a 55 degree temperature, or approximately 5
degree increase. I think I will try this processor in the MSI system
next, to get an idea of how the temp sensor operates.

Cheers,
Dave

Ive heard of others claim big discrepencies but obviously it sounds
flakey on the face of it. I recall way back even then there were
software utils that made your CPU idle much cooler maybe it has some
feature like that.

Frankly Im wondering how flakey these temp sensors are.

So far Ive seen one poster who claims he checked the temps on a
chaintech vnf4 board and they were close at Anandtech -- reasonably
accurate.

Another user at PC perspective - Ive posted his message in another
post , claims its WAY WAY off. Not so much all the time but it seems
to have a ceiling where the actual temp can get way way higher and the
bios keeps reporting the same temps in a narrow range.

Another reviewer doesnt mention temps but he talks about how the
Nforce4 motherboard chips run "red hot" one of my beefs , targets of
suspicion in regards to a lot of claims of data corruption with a
variety of nforce4 boards from different manufacturers.

 

[dave_bonnell]

<snipped>

It is clear to me that temp sensors cannot be trusted. Even if the
sensor is accurate, there are too many other factors involved (sensor
location, orientation, interfacing hardware, software) to throw off the
reading.

My real point in the original post is that my CPU heatsink feels cool,
almost as if it were sitting at room temperature. My (passive)
northbridge heatsink on the other hand, does feel quite warm to the
touch. I was just curious if anyone else had experienced this on a
similar system.

 

[kony]

I found something "interesting" last night, as I built up an Athlon XP
1800+ system. The processor was previously installed on an ASUS board.
Idle temperature was about 55 degrees. I then put some Arctic Silver
5 on, and temps dropped to 49-50 degrees. Note that this reading was
taken from the BIOS....the AsusProbe software consistently reported
about 40 degrees (I think this is socket temperature, or perhaps
something else).

Asus boards were commonly known to report a CPU temp about
5C over actual values, so the current reading is likely a
little high but there's another reason why the difference is
larger than that.

So last night I put the CPU in a MSI nForce2 board. I installed the
same stock HSF. I started the system and proceeded to format a hard
drive, which took about 20 minutes. Then, I rebooted the system,
entered the BIOS, and checked the temperature:

System: 27 degrees
CPU: 19 degrees

WTF?!?! Something's not right here. I do a double-take. At first,
I'm suspicious of the board, which I purchased second-hand. My system
temperature using the ASUS board was only 23 degrees. Clearly the CPU
temperature is not correct....or is it?

It is possible the temp is not currect, but more likely the
Asus board did not implement HALT-Cooling because it's bios
has a chipset bus-disconnect register flipped off. Flipping
it off to prevent the HALT state will be slightly higher
performance for realtime tasks and avoids sudden large
changes in current going from a true idle to loaded state.
These surges can cause instability, if the board or power
supply doesn't respond fast enough.

I "check" the temperature by grasping the heatsink with my fingers. It
is cool to the touch! This is odd, since I'm sure it was warm in the
ASUS system. As a sanity check, I unplug the CPU fan (which, according
to the BIOS is running at 4000 RPM) and watch the temperature.
S-l-o-w-l-y, the temp rises, and I continue to monitor the HSF with my
fingers. By the time the temp reads 30 degrees, the HSF is warm enough
to believe that a processor is actually running (and that the
temperature reading is indeed incorrect).

Ok, it does seem to be cooler in the MSI board. Now it can
be assumed the energy usage was also lower, but so far as
temp goes, it was sufficiently low (enough) on the Asus
board so in the end, it may not matter.

Somewhat confused, I plugged the CPU fan back in and proceeded to
install Windows. Once installed, I let the system run on idle for a
while....and the heatsink is still cool to the touch!

Has anyone else experienced this sort of thing? I upgraded my ASUS rig
to a Bartion 2600, overclocked to 2075 MHz (instead of the normal 1909
MHz). The BIOS indicates a 55 degree temperature, or approximately 5
degree increase. I think I will try this processor in the MSI system
next, to get an idea of how the temp sensor operates.

Your overclocking results (stability) could also be better
with the Asus board because it disables the HALT state,
unless you were to raise the vCore higher in the MSI to
compensate for the momentary voltage drop coming out of halt
state (when CPU comes out, sudden current increase drops the
voltage supply until CPU VRM circuit senses it and
compensates).

It's interesting, but not anything to worry about. Tasks
like video capturing would be best done on the Asus board,
or "maybe" the MSI if it's bios has a setting to disable
HALT-Cooling.

 

[John]

It is possible the temp is not currect, but more likely the
Asus board did not implement HALT-Cooling because it's bios
has a chipset bus-disconnect register flipped off. Flipping
it off to prevent the HALT state will be slightly higher
performance for realtime tasks and avoids sudden large
changes in current going from a true idle to loaded state.
These surges can cause instability, if the board or power
supply doesn't respond fast enough.

That was it. I remember it was either my KT133a or a slightly later
board. They had this thing where it drastically dropped the temps at
idle. On some boards it was turned off cause it caused instability.
There was a whole thing about it but I forget the details it was so
long ago. And there was this software program called Waterfall or CPU
idle or something that supposedly switched it on.

 

[John]

<snipped>

It is clear to me that temp sensors cannot be trusted. Even if the
sensor is accurate, there are too many other factors involved (sensor
location, orientation, interfacing hardware, software) to throw off the
reading.

My real point in the original post is that my CPU heatsink feels cool,
almost as if it were sitting at room temperature. My (passive)
northbridge heatsink on the other hand, does feel quite warm to the
touch. I was just curious if anyone else had experienced this on a
similar system.

it would interesting to see what the differences on both boards are
when you stress the CPU with maybe game playing or some math routine.