I think core temp.... is officially buggered.

[mayakindaguy]

PWNED!!!

lol. Not sure what is causing these discrpencies in temps, but something is definitely off here.

[horrorwood]

HAHAHAHAHAHAHAH sup horrowood, lucky bastage and your 9x multi step 5 e6500. hows that thing clocking btw? yea, im now officially convinced, that right now, something is wrong with coretemp. (and thereby intel analysis tool)

Its stepping 4 actually

Does 3.2ghz on stock volts and 2.4ghz on 1.05

I run 1.15 while at lan cause im not sure on 24/7 stablity only left it a few days at 1.05

[fornowagain]

Go back to basics.

Arctic Silver Thermal Resistance:
0.0045°C-in²/Watt (0.001 inch layer)
Assumed 1in²/ contact area on IHS.

Tuniq Tower 120 Thermal Resistance:
0.16-0.21 C/W (2000 - 1000 RPM)

Using (Intel version):

TR = RCS + RSA = (Tc - Ta)/Pd

where:
TR = Total Thermal Resistance
Tc = Tcase, cpu case temperature
Ta = chassis ambient temperature
Pd = total cpu power dissipation
RCS = thermal resistance, case-to-sink
RSA = thermal resistance, sink-to-ambient

TR = 0.16 + 0.0045 = 0.1645 C/W (Tuniq at 2000RPm with AS5)
Ta = 25C (thermometer inside case)
Pd = 65W (stock settings as Intel TDP ref, 2.4GHz 1.325v, they all say 65W which they can't be, but that's all there is to go on)

0.1645 = (Tc - 25)/65 Therefore Tc = 35.7C



An independent way of estimating the temperature the HSF is actually seeing and proves close to the speedfan/ProbeII values of 34C. If the cpu were at 44C, then at stock speeds and voltages the power dissipation would be Pd = 115W, way more than the stock 65w max thermal design profile.

Going to the other extreme 3.6GHz@1.5v

Using (online version):

Overclocked Watts = Default Watts * (Overclocked Mhz \ Default Mhz) * (Overclocked Vcore \ Default Vcore)²

Pd = 65*(3600/2400)*(1.5/1.325)² = 125W

Therefore Tc = 45.5C



Again very close to the speedfan/ProbeII 46C!!!

What this says is that the junction core temps may well be higher, but who cares? From the calcs the speedfan/probeII readings are accurate. Of course core temps are relevent to internal Thermtrip# throttling. But it's the Tc case temps that's relevent to system builders. Its these temperatures that are used by Thermal Engineers for system cooling solutions. Intel (Appendix D) have extensive details on how it should be monitored for testing with thermocouples. And its the only temperature I have any real control over.

Intel defines a certain Tjunction/TCaseMax temperature for the processor. In the case of Yonah it is 85C° or 100C°. First of all the program reads from an MSR, detects the Tjunction/TCaseMax temperature. A different MSR contains the temperature data, the data is represented as Delta in C° between current temperature and Tjunction/TCaseMax.
So the actual temperature is calculated like this 'Core Temp = Tjunction/TCaseMax - Delta'

I've got some reliable info about where the register that indicates a CPUs TCaseMax is located, reading that register reveals weather its a 85C TCM chip of a 100C TCM chip. I guess most/all desktops are 85C, the mobiles can vary.

One thing I'm not sure about is the TCaseMax (Tc-max) used by the Coretemp (and TAT by the looks of it) Both designed for mobile CPU's with a higher TCaseMax. Is the position of the DTS very near the core surface/IHS interface, otherwise why reference it against the maximum case temp? Or is it a Tjunction register, then is there a Tj maximum?

If I understand it correctly, say that delta is 20C with Tc-max = 85C

Temp reading = 85-20 = 65C

If say Tc-max is 60C

Temp reading = 60-20 = 40C

I'm curious because as far as I can see the TCaseMax for the E6600 C2D is 60.1C, not the 85C Coretemp reports. I sure I'm missing something, forgive me if it's a stupid mistake.

[geforce man]

wow, thats insanely detailed, thanks for the further info, your efforts are appreciated.

[aggybong]

Anyone know what the "VID" value is in Core Temp? Everyone who posts a screenshot has a value of 1.0000v but mine is at 1.0250v. Just wondering what it is :P

[bing]

aggy, your's are hotter, meaning to get up to "stock speed", intel had measured your chip needs 1.0250V while other only at 1 V.

[breakfromyou]

core temp says i load at about 60c. touch my heatsink and its just a little on the warm side. i think its off.

at idle, the temp difference between core 1 and core 2 (of my X2 4400+) is ~7c. at load, they're the same. i dont get it.

[menlatin]

maybe its flaky thermal diode. Intel is coming out with a new stepping for the core 2's. Maybe they caught this.

[PkG.1337]

i m getting 3C on coretemp while using wc on my yonah
(outside temp is 18C)

[The Coolest]

I suggest you carefully reread this post:

Hello...
This quite a surprise that this thread got so big so fast.
This program was originally designed to detect the Core temperature of Yonah processors. In Intel's document it says that Yonah can have a TCaseMax of 100C, as in most mobile processors, this value is usually between 95-100C. Then I've found out that all next gen Core CPUs have the same kind of sensor.
What I also found out is that on B0s and some B1s it acted weird showing all kinds of very low or very high readings.
Regarding the TCaseMax, you have to remember, TCaseMax is the maximum temperature the top of the core can get to, Tjunction is the maximum temperature that the junction of the core with the PCB can get to, and that number is usually much higher.
The fact that throttling doesn't start as soon as you pass 85C doesn't say anything.
I've got some reliable info about where the register that indicates a CPUs TCaseMax is located, reading that register reveals weather its a 85C TCM chip of a 100C TCM chip. I guess most/all desktops are 85C, the mobiles can vary.

Regarding accuracy, the program is only as accurate as the sensor which is reporting the temperature, but then this sensor is a digital sensor so it is not prone to interference. The thermal diode that the motherboard reads is, as there is a separate circuit on the motherboard that reads the diode, which can distort the actual readings and make them very inaccurate, add that to the inherent inaccuracy of all thermal sensors and your temperature reading can be very much off than what it is in reality.
In addition as been said already, the DTS is located in the hottest part of the core, while the thermal diode isn't.

I've written this program according to Intel's datasheet.
It clearly says that the temperature is derived from the TCM value.
As I said it is only as accurate as the sensors so there's no reason saying that Core Temp/other tool isn't working properly because it is, its simply reading a value from a register,

I'm not going to defend my program too much, I'm not forcing anyone to use it, and I understand that when the readings are too much off that you'd want to use a different program to measure temperature. That doesn't mean you have to say its total crap and unreliable, because its more reliable than the kind of software that reads a motherboard's sensor.

Again, I am not sure why Intel chose to calculate the temperature in reference to TCaseMax, but that's the way it is, and this sensor is located in the HOTTEST part of the core, so the high temps some of you see might actually be close to reality.

*EDIT*: I haven't read any updated docs in a while, I might be confused. I will have to recheck and see if the 85C value is indeed the TCaseMax or Tjunction
Either way it still doesn't change anything I've written in this post.

[astaris]

Hello...
The fact that throttling doesn't start as soon as you pass 85C doesn't say anything.

Well i think it says something (from intel datasheet):
"The temperature reported over PECI is always a negative value and represents a delta below the onset of thermal control circuit (TCC) activation, as indicated by PROCHOT#.
Therefore, as the temperature approaches TCC activation, the value approaches zero degrees Celsius. At zero degrees, the TCC activates"
So if you reach 85C and the TCaseMax is 85, the processor should start to throttle, otherwise the real temp is not referenced to that TCaseMax.

[Karak]

I think it's not program's fault, but problem may be located in CPU itself, it can have buggy readings when not stock voltage or even frequency, so if you want to test temp readings try using stock settings, if you still have not realistic results, then you cannot do anything about it.

But it is simple to make sure you have acceptable temperature - I know heatsink isn't same temp as the core, but it when you have good contact is it best way for real temperature measurment - just compare some Pentium D 8xx and Core 2 Duo - you will see that Pentium D is running way hotter. When my Pentium 4 reach 65+ heatsink gets really hot, it can burn my fingers when I touch it for some time. And as far as hot spots are concerned - that doesn't mean anything if CPU have hotspot here or there - heat should be transfered to heatsink same way on Pentium 4, Pentium D or A64 procs... so if you heatsink isn't hot and you are 100% sure of good contact you can forget about fake temp readings.

[ewsforos]

For the love of god people... You can't compare the temps you are getting till now with the temps that come from the die itself... actually, in a sense, ALL programs are right.

Most programs get their temp reading from a thermistor near the die/on the socket, and that's what we've all been accustomed to. Sure coretemp has some problems in some cases (where for example it displays temps below ambient with watercooling etc), but for the cases where it is actually working (reporting realistic temps), pretty much the temps from all utilities are correct.

The temps given from motherboards are usually read from a sensor on the socket... Even if you put one on the IHS this doesn't mean that the temp is that of the core itself since it dissipates on the IHS.

For example, a system will auto-shutdown if the "cpu" temp (see bios/onboard sensor) reaches ~100C as a thermal protection feature. However, parts of the cpu itself, for example where the ALUs are, get MUCH hotter. When the system shuts down at "100C" the actual temp on the die, where the ALUs are located is closer to 120C. This doesn't mean that any of the utilities you use are wrong... on the contrary.. It just depends on what each of us consider to be CPU temp.

Obviously for some time now (YEARS) we've been accustomed to use the thermal probes that are around the socket... However accurate (or not) these might be, it is not the actual DIE temp.

Think of the difference between the two methods as measuring the temp of a car's engine by having a sensor attached on the engine, and by having one INSIDE the engine's cylinder. Obviously, the one inside the cylinder will report MUCH higher temps since it is where the the fuel is burned, but we consider the "90c" in our car's dashboard to be the norm. The engine was build to withstand that internally, but we consider the 90c reported to be the engine's temperature as normal. Now if someone actually moved the sensor inside a cylinder and you saw 400C that doesn't mean it is not normal... It is NORMAL FOR WHERE THE TEMP READING IS TAKEN FROM.

Exact same story goes with the CPU... It is just that someone "suddenly gave as the option to monitor the cylinder's temperature". Sure it's higher, and different but what we used to see, but its absolutely normal. If the temp was read from a sensor at the location we were using till now, it would show lower temps (that's how the difference between coretemp and asus probe for example is explained).

So as you can see, in a sense, they are BOTH right. It all depends on WHERE exactly the temp is read from. Even if you read it on the core, if you read it above the ALUs it will be AT LEAST 15-20C higher than somewhere else on the core. And the core itself is tiny to begin with. So stop fighting over this... Everybody is right It is also wrong to compare previous temps from a sensor probe with ones taken from a die... It could be the exact same thing even though one "is" 15C higher. just read differently... If you want to compare, compare either between sensor probe's temps, or die temps, but do not mix it. This wouldn't make sense. So yes... 80C in coretemp could be the exact same thing as 60C in asus probe as far as the cpu is concerned.

Hope this clears it up a bit

[horrorwood]

Sure coretemp has some problems in some cases (where for example it displays temps below ambient with watercooling etc), but for the most part ALL are correct.

Whats your excuse for mine?

Im using the cooler http://www.tr2tt.com/products/coolers/m18se/m18se.htm

And coretemp says 15 degrees C atm.

[ewsforos]

Whats your excuse for mine?

Im using the cooler http://www.tr2tt.com/products/coolers/m18se/m18se.htm

And coretemp says 15 degrees C atm.

As I said, every piece of software has bugs... It could also be (albeit a remote possibility) a problem with the sensor itself...

However, in cases where ppl see different results like 55C on asus probe/whatever and 70C in coretemp, that's normal and they are both correct in a sense, that's what I'm saying...

In your specific case, obviously something is wrong. I am not in the coding team of coretemp by no means, so I cannot really guess what it could be. It could be a bug, it could be a faulty sensor, it could be different TDP/casemax for your processor, it could be a large number of possibilities that one can only guess towards at this point.



PS: Just realised that what I said was interpreted wrong... I apologise for that... What I mean by "but for the most part ALL are correct" is that it applies on cases where it gives realistic temps... when it is working properly if you prefer instead of giving crazy readings like in your case. Same goes for Intel's utility I'll edit it up in the original post to make it more clear, thanks for that

[horrorwood]

Also what does the VID: mean?

Mines at 1.0125v

[kiwi]

Why do you even care about core temp? As long as everything is working fine, I don't even look at temps Just make sure heatsink contact is good

[Icer]

Regardless if CoreTemp is correct or not, I'm just going to stick with using Probe & AiBooster (as I've done for years) (temp reading seems fine) (both read the same) to read my temps. I can always remember in the back of my mind that my core temps are ~15C higher (not that that helps anything). If I add a new HSF, install a different case fan, or run a benchmark program, it will show me if my temps got hotter or cooler.... that's all I really care about.

[astaris]

*EDIT*: I haven't read any updated docs in a while, I might be confused. I will have to recheck and see if the 85C value is indeed the TCaseMax or Tjunction

Well, it has to be Tjunction because TCaseMax for Intel Core Duo is between 60.1 and 61.4C. Well i think that when the temperature given by the DTS is Tjunction, TC is near TCaseMax. The temperature given by the thermal diode and processed by the motherboard has to be between TC and T_DTS:
TC < Tdiode < T_DTS
So if let's say the temperature given by T_DTS is 85 C, the temperature of the IHS shoud be at 60 C and the temperature of the Heatsink around the base near 50 C.

[GAR]

CORE TEMP is the biggest B/S ever.......i touch my heatipes while priming, core temp says its almosy 70c......i touch it, its barely even warm, not even warm around the cpu.....on my amd system when my cpu was at 60c the heatpipes were hot!....so i dont believe my cpu is running at those temps, therefor i think this program sucks for me and my board.......i have checked my IHS, flat as can be.

[yury2808]

..i touch my heatipes while priming, core temp says its almosy 70c....

same with Zalman 9500 - can't feel any real heat on fins/pipes even very close to cpu...

[mackanz]

Well, the easiest way to see if the HS is acting as it should is just to remove the fan from the HS and look at the digits.

[yury2808]

Well, the easiest way to see if the HS is acting as it should is just to remove the fan from the HS and look at the digits.

ofc, temperature grows around 5C when I down voltage from 12 to 5v on Zalman. the question is - does it work with 100% possible efficience ? why it's so cold ?