Thermometer/temperature probe FOR cpu?

[ntauzin]

What do I use to measure my CPU temp?

 

[blade]

If your motherboard is supported the easiest way is to use motherboard monitor. It's fairly accurate on many motherboards but off up to 6 degrees or more on others.

Most accurate way is to use a thermal sensor. But using those you must be very careful if you have an amd cpu (not sure on newer intel cpu's). The flat end of the sensor isn't long enough to reach the side of the raised cpu core without the plastic covered cord overlapping the cpu. This will cause the heatsink to not fully touch the cpu core because the cord is too thick and a cpu will burn out in seconds. So you'll need to carefully cut off the plastic covering the wire and then place electric tape over it. If you choose this way do be sure the heatsink is fully and evenly on the cpu core before turning the system on.

 

[Mikewarrior2]

Temperature readings on AMD motherboards are not at all accurate... and certainly, 90% of the time not 6C. Trying to justify socket-thermistor readings with those "numbers" is why people don't believe internal diode's are necessary.

Let's look at some examples. Everyone likes to spread that the Asus boards read 8-10C too high. well, comparing asus socket-thermistor to modded internal diode readings are VYW, even the asus boards are inaccurate to 4C too high to 10+C too low.

So even on a motherboard that is too high, there's a 14+C variance in accuracy of its readings, and the readings can't even be predicted to be inaccurate in which direction.

But let's look worst case scenario. With the talk of asus reading 10C too high versus a "normal" board, and we'll use a low reading board for an example: An Iwill KK266 reads about 6-8C lower than Abit motherboards. SO, with a KK266, it is potentialy 8+8+10C off (and yes,if you look at the data, this happens all the time on this motherboard.

so 6C is far too generous to "give" a socket-thermistor in terms of inaccuracy. Especially since, again going to the VYW internal diode versus socket-thermistor tests, there was one very curious result.

He tried both a poor grease application versus a proper grease application, Arctic Alumina, per AS site instructions. The Socket-thermistor picked up ZERO temperature change while the internal diode picked up 7C difference. To replicate this, he tried it again, and managed a 5C Internal diode change but 0C socket-thermistor change. More proof that socket-thermistors are terribly inaccurate, as they can't even pick up small temperature changes.


Mike

P.S. the most accurate reading is a calibrated internal diode. Side-mount thermistors, at best, show 80% of core temp rise over ambient... why? Because you're measuring a mix of side-core, air, heatsink, and grease temps. And each side of a core has different temps. For example, the L2 sides are typically far cooler

 

[blade]

<< and certainly, 90% of the time not 6C. Trying to justify socket-thermistor readings with those "numbers >>



Numbers? I used one "number" and said 6 degrees, not 6c. That's just from my experience.


I agree the side-mounted sensor still isn't the best and most accurate way but for most people that's more than enough. Any other ways are far too much hassle for most to bother with.

 

[fatbaby]

<--lonely post among Elite Members

Try a kingwin multifunction Panel

 

[Mikewarrior2]

Blade,

I have no problem with you posting your own experiences, save one statement you made:



<< It's fairly accurate on many motherboards >>

.

That statement is pure, undeniable FUD. On-motherboard monitoring via AMD motherberboards, regardless of chipset and with a handful of exceptions, is pathetically inaccurate.



Mike