Skuzzy or Bloom25 could u comment on..

[WhiteHawk]

this...  This is utterly contratidictive to what ive been told and heard about CPU temperature...

"The way CPU temperature measured in the BIOS or in the hardware monitor is through an I/O chip on the motherboard. That I/O chip probes a voltage given by the CPU thermal diode, and then BIOS / hardware monitor reads this value and uses some mathematical way to calculate the approximate temperature. According to Intel, there is another temperature sensing diode, which is responsible for the CPU overheating protection. When the CPU junction temperature reaches approximately 135 degrees Celsius (275 degrees F), the processor will turn down itself. That is, we can say the CPU is quite safe if it is less then 135 degrees Celsius (275 degrees F).

Since the temperature is “calculated”, not “measured”, the formula which the BIOS uses will make the outcome different. There is a parameter that is provided by the I/O chip vender, which we expect it to be the standard, so we use this parameter to calculate the temperature of the processor. Once this parameter in the BIOS is changed, the temperature shown in the BIOS PC health or in the hardware monitor tool will be altered. But, in the mean time, the actual temperature of the CPU is still the same. So the temperature you saw might be higher comparing to other motherboard with the same chipset, it is because we use the different parameter to calculate the temperature. We cannot say that we are more accurate, but we sure follow the standard provided by the I/O chip vender, and we believe it should be the proper one."

[bloom25]

It's not quite that simple, but that is essentially true.

The truth is that ANY junction diode can be used as a thermal diode.  A silicon pn junction diode (the most common type) when operating in a forward biased condition develops a voltage across it that changes with temperature.  A typical silicon diode will develop about 0.6V at 25C across its terminals when forward biased.  As temperature increases, that voltage will also increase (if the current is kept constant) by a predictable amount.  The exact increase is computed using the Boltzman constant (k).  This brings up a problem though, as a starting reference voltage must be known for a given current and temperature.  (For example, a standard temperature sensor, the LM335 is something like 2.482 V at 25C and increases 10mV for every degree C increase.)  What this means is that a thermal diode on a motherboard can track changes in temperature very accurately, but unless the reference voltage is correct the exact temperature may not be right.  The motherboard makers probably just obtain a reference voltage used in the bios by sampling a number of boards and taking the average at a known test voltage.  This reference voltage may be different in the software probe utilities, which is why they often read differently than the bios.

[bloom25]

If we are talking about a P4 they can both reduce their clockspeed (called clock throttling) and completely turn off.  135C would definately be the point of turning off completely.  P4s seem to begin to clock throttle between 65 to 70C.

[WhiteHawk]

thnx..  Im top-ending about 59.5C which is pretty close to 60C which is my warning setpoint.  I will up that to 62.5C for a warning and 65C for a shutdown.

[214thCavalier]

My warning is at 85C and shutdown at 90C.
And i reckon i will notice if it clock throttles before that anyway.
It will auto shutdown before harming the CPU at 135C anyway according to Intel who make the cpu.
WhiteHawk i fear you are becoming obsessed with temp figures that are comparable to other Abit users with stock cooling.
If yours were appreciably higher than comparable setups i would worry, but they are not.

Using their measuring figures Abit have told everybody the cpu's are perfectly ok up to a figure of 75C FOR THERE BOARD.
This will not apply to any other manufacturers board and you should not be using their standards for the Abit.

Bloom i recall the IC7-G was actually shutting down many systems because the bios was reading the temps higher than they actually were.
Abit did a bios fix for this (changing the code) if you ask me they just have not changed it enough.
They are sticking with the formula provided by the IO chip vendor when perhaps they should have tested and altered it to bring it in line with other boards.

[bloom25]

From a common sense standpoint the motherboard manufacturers are going to want to be conservative in their temperature measurements.  You definately don't want to risk damage to the system (not to mention lower stability) by reading temperatures too low.

As I said above though, the voltage change on the sensor is very linear.  This means that the temperature may be off, but a 1 degree change on the CPU will still be a 1 degree change on the sensor.  That means that if you think (know) the reading is off by 5 degrees, just adjust your trip points accordingly.

[WhiteHawk]

Originally posted by 214thCavalier
My warning is at 85C and shutdown at 90C.
WhiteHawk i fear you are becoming obsessed with temp figures that are comparable to other Abit users with stock cooling.
If yours were appreciably higher than comparable setups i would worry, but they are not.

 

Ahh, good.  My stock BIOS shutdown was just that so I thought I was good.  But others had suggested that 60C was tempting fate for p4 chips.  Also, the default warning for dr hardware was at 60C so i figured 60C was the safe thing to do.  I can rest easier now.
  Ya, I am starting to catch on to this overclocking culture.  
Buy a short chip, say a p4 2.4 and see how far you can stretch it.
Its kind of strange cuz people will spend more $ to overclock a 2.4 to a  3.2, then they would just buying the damm 3.2.
  I gotta say, its kinda fun:D

[mold]

Originally posted by WhiteHawk
Buy a short chip, say a p4 2.4 and see how far you can stretch it.

Well it depends.  Right now that is a good way to go about it, since most 2.4C's seem to be hitting a good 600+ OC.

Originally posted by WhiteHawk
Its kind of strange cuz people will spend more $ to overclock a 2.4 to a  3.2, then they would just buying the damm 3.2.

Actually, if you just want to get to 3.2, that is not expensive at all.  You can get a cheap mobo, PC3200 RAM, a 2.4C, and $50 on a heatsink, and you will generally get 3.2 with less money than a real 3.2.  The higher cost comes when you try to bring the RAM up along with it.  But faster RAM hugely benefits the overall speed.