Weird temperature readings...

[SiriusStarr]

Hi all,

I recently completed my first build, featuring a Core 2 e6600 and liquid cooling. I plan on overclocking but decided to stability test before I started just to be safe. Anyways, I've had some weird temperature readings and was wondering if you could help me sort it out. The three programs I am using to monitor temperatures are CoreTemp, SpeedFan, and Asus's PC Probe 2. Now, with liquid cooling and no overclock, I was expecting the CPU to run pretty cold, even under load. But when I ran a Prime95 torture test, the CPU temp jumped up to 59 C in PC Probe. Since "hot" for an e6600 is considered to be ~60 C, this is not what I was expecting, as I was expecting some head-room. However, I figured that PC Probe was reporting Tcase, I installed SpeedFan to get the core readings. And here is the weird part. SpeedFan reported identical reports of ~59 C for CPU, but Core 0 and Core 1 were both 52 C!! So I installed CoreTemp, and the temperatures there were also 52 C for both cores. Does anyone have a clue what is going on here? I was under the impression that core temp should be ~5 C higher than cast, not lower. Which is more reliable, core or case? Also, does 52 C seem too high for a non-OC'd e6600 under load core temp, or do I need to reseat the heatsink? (The cooling block itself runs about 36-37 C)

Oh, and idle (in Windows) temps:
CPU: 42-43 C
Cores: 35-36 C
Cooling block: 33-34 C

Thanks in advance for clearing up some confusion...
Sirius

 

[SiriusStarr]

OOPPSSS! Oversight....

<embarassed voice>I forgot to manually set my Vcore...</embarassed voice>

Okay, so that obviously dropped my temperatures a lot, but the weird CPU higher than core remains, and I'd like to know why, even if it's not that critical any more, so if anyone has any ideas, I'm all ears. New readings:

Load:
CPU: 47 C
Cores: 41 C
Cooling block: 34 C

Idle:
CPU: 37 C
Cores: 30 C
Cooling block 33 C

Sorry about the oversight... It's my first time, remember?
~Sirius

 

[CompuTronix]

SiriusStarr, it appears from your use of terms that you've probably read the Core 2 Quad and Duo Temperature Guide - http://www.tomshardware.com/forum/221745-29-core-quad-temperature-guide - however, you may want to take a closer look.


From the Guide:

Section 5: Findings

(A) Tcase is acquired on the CPU Die from the CPU Case Thermal Diode as an analog level, which is converted to a digital value by the super I/O chip on the motherboard. The digital value is Calibrated in BIOS and displayed by temperature software. BIOS can affect the accuracy of Tcase.


Also:

Section 9: Calibrations

Note 1: PECI (Platform Environmental Control Interface) is a BIOS feature on some (not all) recent motherboards which determines the method by which chipsets interpret and manage CPU temperature. When enabled, thermal accuracy is enhanced, and if disabled, temperatures are typically inverted, where Tcase is higher than Tjunction.


And:

Section 10: Results and Variables

Erroneous BIOS releases and temperature utilities are much too often the source of thermal inaccuracies.


Additionally:

Section 15: Troubleshooting

(C) Any hardware and / or software may misreport Tcase and / or Tjunction temps.

(D) BIOS updates will affect the accuracy of Tcase, but will have no affect on Tjunction.

(E) If Tcase is higher than Tjunction, then enabling PECI (if equipped) in BIOS may correct inverted temps.


Regardless of whether your motherboard's BIOS is equipped with PECI or not, CPU temperature (Tcase) basic accuracy is dependent upon processor model-specific BIOS thermal algorithms which are frequently mis-coded. I have noted on numerous occasions where 4 consecutive BIOS updates on the same motherboard / processor / cooler combination indicated CPU temperatures which were anywhere from 15c below ambient, to a believable temperature, to 10c above ambient.

If you follow the Calibration Section in the Guide, you can use SpeedFan to fine-tune your temperatures so that you'll be confident that they're accurate.

Hope this helps,

Comp

 

[SiriusStarr]

Yeah, I overlooked that section... Well, PECI is already enabled in my BIOS, so that's not the issue. From what I'm hearing here, Tjunction is the more accurate of the two, and thus I should probably go by that value, i.e. keeping it below 60-65 C? Can I just go based on the core readings and ignore the Tcase?

Once again, I apologize for my noobishness,
~Sirius

 

[CompuTronix]

From the Guide:


Section 5: Findings

(C) Tcase and Tjunction are both acquired from Thermal Diodes. Tcase and Tjunction analog to digital (A to D) conversions are executed by different devices in separate locations. BIOS Calibrations, factory DTS Calibrations and temperature software are frequently inaccurate.

(D) Intel shows Maximum Case Temperature (Tcase Max) in the Processor Spec Finder, which is the only temperature that Intel supports on Core 2 desktop processors. Ambient to Tcase Delta has known Offsets which vary with power dissipation and cooler efficiency, and can be Calibrated at Idle using a standardized Test Setup.


You can have the most accurate CPU and Core temperatures specific to your rig if you just follow through with the Calibrations Section, rather than depend on the canned values of popular utilities.

Comp

 

[SiriusStarr]

Is there no way out for the lazy... DD

Okay, I ran through the calibrations, and they make me rather nervous.

My Tcase at idle was 43, ambient was 19.9C. X=3 and Y=2, so Z=2.5=3. I end up with an offset of 20.1 C, making my idle temperature 23 C. That sounds waayyy to low for me, and I really don't want to burn out my chip, so am I doing something wrong here? What should Tcase for an e6600 be (roughly) idling? Don't have time to run the calibration under load, but my current core temps are about 13 C higher than case, when it sounds like they should be about 5. Is it possible for sensors to come this far out of whack?

Once again, I stumble over myself apologizing for my relative newness on the subject.
Sirius

 

[CompuTronix]

If you've followed the Test Setup to the letter, and if your E6600 Spec # is SL9Z, then you're Idling at 12 watts, and your CPU Idle temp is appropriate. If your E6600 Spec# is SL9S, then you're Idling at 22 watts, so your CPU temp should be calibrated at least 3c higher. Ultimately, the CPU temp is as accurate as the measured ambient, and the cooler value.


Section 6: Scale

Since Tcase indicates CPU Die temperature only, it is not used for Throttle or Shutdown activation, however, as Tcase Max will be exceeded before Tjunction Max is reached, Tcase Max is always the limiting thermal specification.

Scale 5: Duo
E6x00: Tcase Max 60c, Stepping B2

-Tcase/Tjunction-
--60--/--65--65-- Hot
--55--/--60--60-- Warm
--50--/--55--55-- Safe
--25--/--30--30-- Cool

 

[SiriusStarr]

I think the problem here is that I have a liquid cooled system, which means that at idle, my CPU can't cool to ambient + (ambient Tcase delta). It has to cool to the average of all the components in my system. Because my NB and GPU (i.e. much hotter components) are also liquid-cooled, even at idle the cooling blocks themselves don't get below about 34-36 C (based on two independent thermal probe readings). Thus I don't think my CPU can ever get below about 34 C, since it's not going to start cooling the cooling blocks themselves.

Does anyone know how to calibrate temperatures for a completely liquid-cooled system? I don't see how I really can reach the lowest idle temperature. Can I just assume that Tcase (at idle) reaches the same temperature as the water-block plus a degree or two?

Thank you for all your help on this CompuTronix; my rig is just making it rather difficult... :-(

 

[CompuTronix]

Thanks for clarifying your cooling loop. Is the CPU cooling block first in the loop before the NB and GPU? Also, a correction from your previous post; Y = 1 in this instance, since liquid cooling is considered high-end. So for your rig, CPU cooling block +2c may be appropriate for Tcase Idle. Under the conditions of the Test Setup, if your ambient is consistent at 20c, and the CPU cooling block stabilizes at 33c, then Tcase Idle should be 35c. Since the Cores obviously can not Idle below the CPU cooling block, or Tcase, the only remaining blank to fill in is Tjunction Load.

Comp

 

[Lupiron]

LOL, nice. That did help alot, and all I did is read it all just now!

I tried that as well. I found that you either have to fork over alot of money for real good components, of leave the CPU on its own water block! You are right, the ambient of your liquid depends on everything else in the loop, and then the cooling radiator type and fans.

I had an ultra or two on the loop, and that was a big no go, as the ultras idle way higher normally. 50c. So that made my liquid idle pretty high off the top, and the warmer it gets the lower its efficiency gets.

Glad I dont care about cooling, only about speed! I'll leave that for CompuTronix!

Of which I have a question for! I am sure its prolly posted somewhere, but I have been known to be lazy! Whats up with the "usual" progies showing one core colder than the other, or a pair of cores and such? Is that due to those calculation errors for the junk that translates them to a temperature?

Or are the cores simply better placed under the IHS and therefor shows one or two as cooler or hotter than the others?

--Lupi!

 

[CompuTronix]

Lupi, good timing. The answer to your question is "both". I just posted the following on another thread: Q6600 OC and temp question - http://www.tomshardware.com/forum/246773-29-q6600-temp-question


"jhbwolverine, although Cores 0 & 1 normally differ a few degrees from Cores 2 & 3 due to factory DTS calibrations, your 10c deviation between highest and lowest Cores are twice what is typical. You may want to inspect your processor's Integrated Heat Spreader for flatness, as well as the CPU cooler's Heat Sinc. A combination of imperfections can result in an area of minimal contact above a particular Core, which can result in uneven temperatures."


Hope this helps,

Comp

 

[Lupiron]

Hehe, I do that, but I was meaning the more normal 3 or 4 c thats almost always there. Is that just due to the location of the cores and the physical characteristics of the IHS and chip design, or can it be blamed on a miscalculation of calibration software?

Any clue?

--Lupi!

 

[CompuTronix]

All temperature softwares read the DTS (Digital Thermal Sensors) which are calibrated at Intel's fabrication facilities, so if Intel mis-calibrates the Cores, (which happens all too frequently), then that's what you'll see. Also, remember that a Quad Core is composed of 2 Dual Cores, which may have differing DTS calibrations.

 

[Lupiron]

Yup! There is one guy who has cores 10c apart, lol! Thanks! My biggest spread on all 9 of my q6x00s was 6.

--Lupi!

 

[CompuTronix]

I've tested Quads which had an individual Core that indicated 6c below ambient, as well as an individual Core that indicated 39c higher than it's neighboring Cores, and ran just fine without prematurely throttling. Also, consider all the DTS problems currently being reported by users of 45 nanometer E8xxx and Q9xxx processors. Intel sometimes doesn't get it right.

 

[SiriusStarr]

The CPU is first in the loop, kind of. The liquid comes in and splits to two loops, one goes to CPU, the other to GPU, so yes, the CPU is getting coolant straight out of the radiator. I have yet to see more than a 1 C difference between cooling blocks though. (My system has a pretty darn high flow-rate...)

Thanks! I'll use the CPU block + 2 C for the calculation. That should be close enough. I'm not real concerned about knowing my temps to the degree. I just want to make sure my Tcase stays below 60 C under load, so I know how far I can overclock. I'm at 400x8 and my Tcase (using the above formula) is about 52 C under Prime95 highest heat, so I'm pretty happy with that.

Thanks again,
Sirius

 

[Lupiron]

LOL, thats very true. I kinda figured you had to use a bit of common sense and assume something is being miss read, especially if you are sure the cooler is seated properly.

I did notice that on a few of them, the temps even out at the higher degrees. Like they all would read 57c during small fft testing. I'm always shocked when I see them all exactly the same when fully loaded, hehe. Its not natural! (Thanks to whatever various array of errors that happen to be plaguing your chip. There are 197 known erratas, and only a few have been corrected!)

Thanks again! Will be watching whats going on!

--Lupi!