temp probe under CPU vs. probe next to die.

[samsonx]

I've heard that with xp's the diode in the socket is incorrectly reads the temp do to the resistors on the botton of the chip.

What I'm wondering is the diode reads higher or lower than the actual die temp.

[Random Nonsense]

the in socket thing i have seen do both

on my XP1800+

Socket: 31
Die : 65

Socket:30
Die : 12

spode also saw the second one with his own eyes.

the thermocouple next to the core is the second best way of measuring temps on an XP, ideally you want to make an internal diode reader, but its kinda over my head so cant explain that to you. sorry, i know some of the guys here have VERY good knowledge of this kind of thing, with any luck someone will spot this and give you info on it all

[cloasters]

The in-Socket thermistors are said to be off by 5C--as a general rule. Higher or lower than the real value? The only way to know is through experience. Arrgh. I think that the Asus A7S-VM's thermistor reports a value that's 5C higher than reality.

I read that no manufacturer has built a mainboard that accurately reports the Athlon XP's(nor Morgan-cored Duron's) internal temperature correctly. Yet. It's about freakin' time, gentlemen!

[SPQQKY]

It's not the mainboard manufacturers fault, the AMD chips do not have internal sensors like the Intels. They can only do their best by placing a sensor under the chip. I use a Digital Doc 5 to monitor my XP's temps. I place the sensor next to the core and get readings as accurate as one can expect. Currently, my XP is running at 34.9c.

[Shobiz94]

The AMD XP/MP and the new Durons have an internal temperature sensor, just no motherboard supports them as of yet.

[SteveI]

[Aedan]

Originally posted by samsonx
I've heard that with xp's the diode in the socket is incorrectly reads the temp do to the resistors on the botton of the chip.

What I'm wondering is the diode reads higher or lower than the actual die temp.

As you probably know, the thermistor mounted under the socket reads the temperature of the packaging. It takes time for heat to travel through the packaging, so the temperature you read can be seconds behind the temperature of the die.

In addition, as the packaging has thermal resistance, the temperature of the bottom of the packaging is going to be cooler than the die itself, as heat is lost through the PCB to the other side of the board.

Put this together and you'll find that the thermistor will read a temperature that's an average of the last n seconds less a bit for the heat lost!

If I wrote a program that maxed out the CPU for half a second, then idled the CPU for half a second, and repeated, then the core would heat up and cool down rapidly, but the thermistor would report the average between the hot and cold temperatures!