AMD CPU temps poll!

[Mikewarrior2]

Sleater,

motherboard manufacturer compensation. this varies between manufacturers in an attempt to "correct" the readings and give better representation of "die" temperature.



Mike

 

[Jerboy]

CPU temperature is rather meaningless unless the motherboard industry has the thing calibrated to show CPU die temperature. Otherwise it doesn't mean a jack other than relative comparision within your system.

With thermal probe set at same parameter, XP will appear to be running at lower temperature since its composite packaging is less thermally conductive than Thunderbird's ceramic. It also depends on whether or not thermal probe is touching the bottom of the processor, contact pressure, and linearity and repeatability of the measurement setup.

Some motherboard will detect CPU type and automatically adjust the temperature sensing mode, some requires jumper(Asus A7V266 and such) and others simply don't compensate resulting in lower temperature reading and excessively happy operator.

Delta T between core and bottom is not always equal and calibration is more compelx than it seems, therefore accurate temperature measurement is not possible without using on die thermal diode or elaborate calibration procedure. (ever guessed how much mobo will cost if + - xx degree was part of the spec it and calibrated one by one like lab equipments?)

 

[Jerboy]

TTT

 

[cookieman]

Hi~

It seems to me that Mike is trying to explain an important thing here. But no one is listening...

The ambient temp would be a must to know in these comparisions. The type of MBs taking part in the comparision is also very important.
But this is not enough: a lot of other factors can make a big impact in the temp readings
- is the CPU at full load or idle (some ppl do not know what is full load). Speed of the CPU is also useful...
- is the case open or close when tested.
- the CPU HSF are not the same so some of them actualy cool the termistor better then the others.
- is the termistor is making proper contact with the CPU back?
- case form and case airflow can deviate the reading the termistor readings (cooling the termistor more).
- is the case on the ground or is it on the table (the air near the ground is cooler you know).
- different bios revisions make different temp corrections.
- some ppl have hot Videocards near the CPU's HSF.
- some have posted other factors that can modify the reading.

You can always aleviate some of the inacuracy factors I just wrote and you will be a little closer but even that will not be too informative
to other MB brands or configs. But just asking for some random temps is plain useless...

From time to time there are a lot of threads that discusses "Is 50C too much?". It seems to me that those ppl have no clear idea of termodinamics and cannot make any bad/good guess of their own CPU core temp.

rogue1979: I always "ignored" the reports of ppl claiming to achieve great OC without suplying the data you just wrote down. Some of them are reporting 39C temps at full load thought.... And that do not seems fishy to them.

You ppl should go read the stuff on Mike's page to know your temp limits. That is far much informative than this.

Cheers,

 

[Mikewarrior2]

Ha, straight from the horses (amd's) mouth in this thread: JCMoran's E-mail reply from AMD regarding temps.

to quote the other thread specifically:

The "normal operating temperature" of a system or a processor is a difficult question to answer. What might be "normal" for one system will not be normal for another. There are many variables in a system that will change the normal operating temperature. Case size, installed components, airflow, processor speed, power supply, voltage settings, heat sink, thermal interface material, workload, ambient air temperature, etc. all play a role in the normal operating temperature for your system and processor.
Also note that the accuracy of the thermal probe, hardware monitor and conversion from an analog signal to digital also impact the actual temperature that is being reported in the BIOS or software.
Therefore, we cannot say what should be "normal" with respect to the operating temperature.



Mike

 

[Mikewarrior2]

bump so people can see amd's view of "normal" temps.

 

[Mikewarrior2]

bump again!


(what, no one wants to reply to "amd's response"... ).

 

[Rack]

Just for you, Sleater. Here are my graphs on my a7v and my 1GHz morgan at 1160 and 1.85V:

mc370
dd new maze2/eheim 1250/dd cooling cube

 

[NesuD]

Mike these threads aren't useless step back and look at this one for example. You flew in here and gave everyone the truth about socket A temp readings. How is that useless? If these threads get you in here to keep spreading the word than at least that one useful thing has come of it. By the way everyone for what it's worth right now i am reporting a cpu temp of 44c with a motherboard temp of 24c running my cpu at 1.47 ghz with a core voltage setting of 1.88volts on an MSI K7T turbo R.

P.S. notice how i qualified all that with "for what it's worth".

 

[SpeedTester]

I agree Nesud, 131 peeps have put in their temps so obviously some of us
find this useful.

 

[Mikewarrior2]

Speedtester,

Nice way of taking Nesud's post and using it to quantify this craptacular thread. 132 people posting temps does not make it useful, or even accurate.

Once again, since no one wants to believe me, i will give you AMD's take on "normal" temps:



<< The "normal operating temperature" of a system or a processor is a difficult question to answer. What might be "normal" for one system will not be normal for another. There are many variables in a system that will change the normal operating temperature. Case size, installed components, airflow, processor speed, power supply, voltage settings, heat sink, thermal interface material, workload, ambient air temperature, etc. all play a role in the normal operating temperature for your system and processor. >>



That's an e-mail response from AMD... take it however you like to, but the fact is, too many variables need to be addressed before making any kind of temp "poll" useful.


Mike

 

[NesuD]

Ok mike lighten up my post was simply a little lighthearted razzing. I have seen the amd letter on socket a temp readings before and you know that you will get no argument out of me on that subject. All I was saying is while the posted point of this thread is useless which by the way i agree with you on. It does provide the opportunity for you to enlighten some of the less knowledgeable around here. That and only that was what i was saying.
Actually Speedtester I didn't put my temps into the poll as there was no category for my motherboard, hsf, case, and fan configuration and my cpu doesn't have an onchip diode which my board couldn't read if it did. My Temp posting was simply to demonstrate how useless comparison is since you must also accnt for several variables like the board used to take the temp, the hsf the quality of the case ventilations. basically the variables are so wide ranging that unless you have the exact hardware configurations side by side taking the temps at the same time they are pretty much useless for comparison purposes and definately useless for monitering actual cpu die temp.

 

[The Sauce]

Has anyone noticed that we are getting a very nice, bell-shaped standard distribution of temps in the poll? Although our (n) is still pretty small, any student of statistics will know that this in itself denotes some degree of consistency in the data collection. If, as Mike suggests, there were a few factors significantly influencing these temp readings (which I am not arguing), you would have expected to get a multimodal distribution...but there it is...a nice bell shaped curve. Kinda tough to explain, eh? Maybe this "craptacular" thread has some legitimacy after all...we shall see.

 

[RumXZ]

on idle my cpu usually runs at under 50 degrees C, but on load its usually over 50
got a duron 800 o/c 896 mhz 8x multiplier 112fsb and 1.65v

have a stock HS with a TT golden orb fan

 

[Richardito]

<< Sure it is useful... to create confusion among those mb's whose temps are normally hotter.

Mike >>



Man, take a chill pill... The world isn't a perfect place and you don't have the monopoly on thermal discussions.

 

[Haden]

I've got 35C Idle, 45C full load on 1500+@150x10 (volcano 6Cu), doesn't seem very accurate to me