Temperature and VCore issues and observations

[Guest]

You folks give some good leads and insights into cpu and MB
temperatures and measurements. I have an additional question and some
observations...share the observations and ask a question or two. I am
using the latest versions of Probe, MBM5, and Bios.

I have a replacement A7v333 rev1.04 board running an AMD XP 1600+ that
apparently has a bogus Vcore reading...even with a bios update, which
was recommended by Asus 2nd level tech support...using either jumper
settings or jumperless settings. However, this appears to be a bogus
reading since the cpu temperature stabilizes pretty quickly. But
stabilizes at what? This is the questions I was looking for an answer
for. At a vcore voltage of 2.04 when it is jumpered for 1.65-1.70, my
cpu should be frying...but it isn't.

At room temp...all off over night...cpu, MB, bios temps via Probe,
MBM5, and bios all show 21C. Not bad, but what about linearity when
temps go up? This is one observation I would share. It is trivial
circuitry to linearize readings over such small ranges as 50C, so my
guess is that linearity is not a true issue. The
diodes..thermocouples... are inherently reasonably linear and
correctible through calibrated tables over this small range, diode to
diode variability included.

Second observation. The board mass and heatsink mass mean that over a
minute or two, once temps have stabilized, they will stay the same
during reboot and sensor cycling...going from one monitoring method to
another....that seems to be a point no one points out when dealing
with non-simultaneous measurements...over very short times it is
inconsequential what changes occur.....mass that heats up over time
takes time to cool off..likely longer since the heat source is a
forcing source and the cooling is more passive. Only the naked
diode...not in contact with any other mass...would heat or cool
quickly.

Now, I do have a question. Once the system temps stabilize, MBM5,
Probe, and the bios give temps that are consistent and within a degree
of oneanother. The MB...called the case, I think, in MBM5... is about
30C, the cpu diode in MBM5 is about 35C and the socket...whatever this
is...is 47C. Now, where is the diode relative to the socket? The
sensor closest to the cpu should be the hottest..is this the diode?
The socket should be next..unless the temperature sensor is adjacent
to the socket but farther from the cpu. And then the question of where
is the MB sensor relative to the socket? Under loads, these only go up
a few degrees C. Hence my assumption that vcore is bogus and really
running where I set the jumpers at 1.65v.

The interesting thing is another system with an A7N8X Deluxe board
running an AMD 2200+ XP cpu runs almost cold. But some sensor
measurements are missing from MBM. The case is 30C and socket is 36 C
but the cpu diode is zero. Probe shows the cpu at 32 and MB at 29C.
The bios has the MB at 30 and the cpu at 40C. So, what the hell is
what and what happened to the diode? The only thing I see in all this
is that this system is running way cold but very stable in the turbo
mode....which, in the long run is just what I want. I just don't like
the ambiguity.

Any dialog on these issues would be appreciated.

Henry LaMuth

 

[Ed]

You folks give some good leads and insights into cpu and MB
temperatures and measurements. I have an additional question and some
observations...share the observations and ask a question or two. I am
using the latest versions of Probe, MBM5, and Bios.

I have a replacement A7v333 rev1.04 board running an AMD XP 1600+ that
apparently has a bogus Vcore reading...even with a bios update, which
was recommended by Asus 2nd level tech support...using either jumper
settings or jumperless settings. However, this appears to be a bogus
reading since the cpu temperature stabilizes pretty quickly. But
stabilizes at what? This is the questions I was looking for an answer
for. At a vcore voltage of 2.04 when it is jumpered for 1.65-1.70, my
cpu should be frying...but it isn't.

At room temp...all off over night...cpu, MB, bios temps via Probe,
MBM5, and bios all show 21C. Not bad, but what about linearity when
temps go up? This is one observation I would share. It is trivial
circuitry to linearize readings over such small ranges as 50C, so my
guess is that linearity is not a true issue. The
diodes..thermocouples... are inherently reasonably linear and
correctible through calibrated tables over this small range, diode to
diode variability included.

Second observation. The board mass and heatsink mass mean that over a
minute or two, once temps have stabilized, they will stay the same
during reboot and sensor cycling...going from one monitoring method to
another....that seems to be a point no one points out when dealing
with non-simultaneous measurements...over very short times it is
inconsequential what changes occur.....mass that heats up over time
takes time to cool off..likely longer since the heat source is a
forcing source and the cooling is more passive. Only the naked
diode...not in contact with any other mass...would heat or cool
quickly.

Now, I do have a question. Once the system temps stabilize, MBM5,
Probe, and the bios give temps that are consistent and within a degree
of oneanother. The MB...called the case, I think, in MBM5... is about
30C, the cpu diode in MBM5 is about 35C and the socket...whatever this
is...is 47C. Now, where is the diode relative to the socket? The
sensor closest to the cpu should be the hottest..is this the diode?
The socket should be next..unless the temperature sensor is adjacent
to the socket but farther from the cpu. And then the question of where
is the MB sensor relative to the socket? Under loads, these only go up
a few degrees C. Hence my assumption that vcore is bogus and really
running where I set the jumpers at 1.65v.

The interesting thing is another system with an A7N8X Deluxe board
running an AMD 2200+ XP cpu runs almost cold. But some sensor
measurements are missing from MBM. The case is 30C and socket is 36 C
but the cpu diode is zero. Probe shows the cpu at 32 and MB at 29C.
The bios has the MB at 30 and the cpu at 40C. So, what the hell is
what and what happened to the diode? The only thing I see in all this
is that this system is running way cold but very stable in the turbo
mode....which, in the long run is just what I want. I just don't like
the ambiguity.

Any dialog on these issues would be appreciated.

Henry LaMuth

There is a thermal diode in the center of the CPU socket, this is where
Asus PC Probe gets the CPU temp. (All Socket-A boards have a diode here)

Using an A7N8X mobo and MBM5 you can read the temps from the thermal
diode built into the Athlon XP CPUs and the thermal diode in the CPU
socket.

Set up MBM something like this...

Asus 1..................Case/Mobo (Rev < 1.06)
Asus 1..................Socket (Rev 1.06 and up)
Asus 2..................Socket (Rev < 1.06)
Asus 2..................Case (Rev 1.06 and up)
W83L785TS-S Diode.......CPU Diode (in most cases)
W83L785TS-S.............CPU Diode (in few cases)
Asus 3, Asus 4..........Unused/Optional Self-Installed Sensor

Fan Asus 1..............CPU Header
Fan Asus 2..............Chassis Header
Fan Asus 3..............Power Header
Voltage Configuration...ASB100 Standard 2


As far as voltages go, get a voltage meter and put the ground lead to
the case and then probe around with the other lead to the center of the
MOSFETs (voltage regulators) to get a more accurate reading of your
+3.3V +5V, Vcore, etc,etc..

Cheers,
Ed

 

[P2B]

Ed wrote:

[snip]

As far as voltages go, get a voltage meter and put the ground lead to
the case and then probe around with the other lead to the center of the
MOSFETs (voltage regulators) to get a more accurate reading of your
+3.3V +5V, Vcore, etc,etc..

Probing the MOSFET drains will give misleading results. Onboard
regulators are normally either Pulse Width Modulators (used for 3.3v and
Vcore which draw significant current) - with two MOSFETS connected
source-to-drain on the output side - or Linear Controllers (used for low
current requirements like the GTL bus and clock drivers), which usually
do not use MOSFETS. Probing the PWM's upper MOSFET drain will typically
read the unregulated input, while the lower drain is regulated but
unfiltered output.

For accurate readings, you should measure after the filter capacitors.

P2B

 

[Ed]

Ed wrote:

[snip]

As far as voltages go, get a voltage meter and put the ground lead to
the case and then probe around with the other lead to the center of the
MOSFETs (voltage regulators) to get a more accurate reading of your
+3.3V +5V, Vcore, etc,etc..

Probing the MOSFET drains will give misleading results. Onboard
regulators are normally either Pulse Width Modulators (used for 3.3v and
Vcore which draw significant current) - with two MOSFETS connected
source-to-drain on the output side - or Linear Controllers (used for low
current requirements like the GTL bus and clock drivers), which usually
do not use MOSFETS. Probing the PWM's upper MOSFET drain will typically
read the unregulated input, while the lower drain is regulated but
unfiltered output.

For accurate readings, you should measure after the filter capacitors.

P2B

Cheers,
Ed

Right, but even before any filters you should still get a better idea if
his reading of 2.10v for the vcore is really any where near that
voltage.

I've checked 3 of my boards at the Mosfet and 2 of them show the
voltages exactly as I see them in the BIOS/Heath Monitoring programs,
the 3rd board only shows a 0.5v difference at the most.

Cheers,
Ed

 

[Guest]

Ed wrote:

[snip]

As far as voltages go, get a voltage meter and put the ground lead to
the case and then probe around with the other lead to the center of the
MOSFETs (voltage regulators) to get a more accurate reading of your
+3.3V +5V, Vcore, etc,etc..

Probing the MOSFET drains will give misleading results. Onboard
regulators are normally either Pulse Width Modulators (used for 3.3v and
Vcore which draw significant current) - with two MOSFETS connected
source-to-drain on the output side - or Linear Controllers (used for low
current requirements like the GTL bus and clock drivers), which usually
do not use MOSFETS. Probing the PWM's upper MOSFET drain will typically
read the unregulated input, while the lower drain is regulated but
unfiltered output.

For accurate readings, you should measure after the filter capacitors.

P2B

Cheers,
Ed

Right, but even before any filters you should still get a better idea if
his reading of 2.10v for the vcore is really any where near that
voltage.

I've checked 3 of my boards at the Mosfet and 2 of them show the
voltages exactly as I see them in the BIOS/Heath Monitoring programs,
the 3rd board only shows a 0.5v difference at the most.

Cheers,
Ed

Under most circumstances, a vcore as high as 2.1 volts would basically
be frying the cpu...but I am not getting that. In fact, it is running
fairly cool at 1400MHz...basically what it would be rated for if it
were actually at the 1.65v the jumpers are set at. So, why wouldn't
this be a bogus bios or mbm5 or probe measurement? Bad sensors are
fairly common occurances.

My alternative is to send the board back regardless...a bad sensor is
the same as a true voltage problem. Something is wrong and cannot be
fixed by simple processes. I could measure the actual voltage, but in
this case, why bother? Either the cpu fries or the board is OK as
is...which is what it seems to be, which is OK. My inclination is to
get a new board from Asus if I can.

Henry