concerns about temp difference between Tcase and Tjunction

[cskamil]

Hi,

I'm planning to overclock my Q6600 but just can't start. I just found this thread: http://www.tomshardware.co.uk/forum/221745-11-core-quad-temperature-guide

I read and try to understand that line by line. I also read the responses to that thead and searc other threads to find out a solution but couldn't find a promising answer.(Just questions )

Q6600 is installed @ 2.4GHZ on GA-P35C-DS3R - rev1.1 and with a NOCTUA-NH-U12P CPU cooler. According to thread I found my Tcase Idle which is 34c - 35c. Ambient inside chassis( CoolerMaster Cosmos 1000) is 29c. I guess this is an expecting value. My concern is Tjunction(Core) temperatures shown by speedfan 4.35. After 10min torture test on Prime 95, Tcase is 54c-55c and my cores temperatures:
Core 0 : 69c
Core 1 : 69c
Core 2 : 66c
Core 3 : 62c
( I also get similar readings @ TAT)

In the thread it is said that the difference between Tcase Max ( 55c) and Tjunciton (69c) should be 5c. But it is about 14c. Is it reasonable to set offset in speedfan in respect to Tcase reading only? Even if I set that offset on speedfan, I would get same greater temperature readings on TAT.
I reinstalled my cooler 3 times with using zalman ZM-STG1 and Noctua NT-H1 thermal grease but no difference on temps. At realtemp 2.70 I observe temps closer to Tcase( CPU) temperature on speedfan but again huge difference with core temperatures on speedfan and TAT. Why there is huge difference between Tcase and Tjunction?

Thanks for your help and I would be sorry if someone had answered this before cause I could't find it

 

[Mondoman]

Use realtemp and ignore speedfan and TAT.

 

[CompuTronix]

Thanks Mondoman,

Good advice about TAT, but some users want more temperature information than just Core temperatures, especially since Intel dropped the ball at the recent IDF 2008 DTS (Digital Thermal Sensor) Presentation. As you know, my Temperature Guide is based on SpeedFan, because CPU temperature (Tcase) and Core temperature (Tjunction) can both be calibrated, so both sets of temperatures can be used to cross-reference one another for accuracy.

As Intel now admits to the severity of DTS "Slope Error" for monitoring Core temperatures, the author of Real Temp has already adjusted his calibration values, which now match SpeedFan's values. Regardless, SpeedFan has always provided CPU temperature (Tcase), while Real Temp does not. Additionally, Intel's Processor Spec Finder - http://processorfinder.intel.com/Default.aspx - has always used CPU temperature for the Thermal Specification, not Core temperature, due to "Slope Error".

Finally, since my calibration procedures for SpeedFan do not use the hotly debated and still partially undisclosed Tjunction Max values, the resulting accuracy, when calibrations have been completed, are not affected by these values, which Intel admits vary from part to part, and are not recommended for temperature monitoring. This is why Tjunction Max values are round figures such as 100c, while Tcase Max values are listed to the tenth of a degree, such as 72.4c.

Comp

 

[CompuTronix]

cskamil,

Regardless of your Tcase / Tjunction calibration problem, your temperatures are far too high at stock.

(1) Have you checked your Q6600 IHS for flatness? Is it concave or convex?
(2) Is your Q6600 Stepping G0 or B3?
(3) What is your ambient as defined in the Guide?
(4) What are you using as a measuring device, and is it accurate?
(5) Did you follow all Items in the Test Setup to the letter, with case covers removed and all fans at 100% RPM?
(6) Have you actually performed the calibrations?
(7) Since Section 8 in the Guide doesn't include TAT, then you shouldn't use it.

TAT @ 100% Thermal Load (which loads all registers with all 1's) = Prime95 Small FFT's @ 114% Workload. Therefore, TAT's temperatures run ~ 5c higher than Prime95, and it's curved for notebook processors, so TAT's temperatures are unrealistic and inconsistent with Prime95, which replaced TAT in the Guide nearly a year and a half ago.

I keep the Guide updated with the latest information and calibration procedures, so it's current at this time. As the Guide became a Sticky 2/2/07, the first posts are quite out of date. Also, since I've always asked that users post their questions in new threads, this is why my answers don't appear in the Guide's thread.

Comp

 

[cskamil]

Thanks CompuTronix,

(1) I guess I need to check the surface of my CPU. I didn't check it and I don't know whether there is another method or not.
(2) Its revision is B3 from CPU-Z
(3)- (4) I guess the room temperature is around 25c- 26c. I have bougth a new chassis fan which is Thermaltake Cyclo 8cm logo fan. I'm using this fan as a measuring device. It has a temperature display on it. It's not located at exaust. It's top of air intake which is from the bottom of the chassis. It's showing 29c - 30c. I'm not totally sure about its accuracy.

(5) I attach all my fan's cable directliy to my PSU which as far as I know should make all fans at 100% RPM. I also check the BIOS settings. I also calculated the Tcase as Ambient + Z. I calculate Z as (6+3)/2 = 4.5c ~ 5c. The fan thermometer shows 27c inside the case and with all covers removed. I guess the idle cpu temperature is plausible. I performed test again after removing case covers. Idle temperatures are at speedfan:
Tcase: 32c ( which is shown as Temp2 in speedfan and rename it is declared in guide with 10 seconds torture tests )
Core0: 45c
Core1: 44c
Core2: 46c
Core3: 40c
After 10 min torture test temperatures are:
Tcase: 50c
Core0: 65c
Core1: 63c
Core2: 61c
Core3: 57c

(6) I didn't performed calibrations since I'm a bit confused about using Tcase to set offset of Tjunctions because as I said there is huge difference betwwen them.
(7) Ok I won't use it. But if I set for eg. core0 offset as '-10' which will cause core0 temperature shown as 55c and the diffrence became 5c. But in the reality the core0 temperature continiues to be around 65c.

Since the temps are high at my stock fan and previous case, I decided to buy a tower fan which is NOCTUA-NH-U12P and a new case which is CoolerMaster 1000 to increase the airflow and the cooling to perform OC. But I guess as you said these temps are higher than other Q6600 users and I feel that I need help because I'm newbie to OC. This is my first time to install a tower fan but I reinstall it again and again with using new thermal grease but reinstalling just causes minor differences or no difference. I touched my tower fan after torture test, It was warm not hot.

Thanks for your responses.

 

[CompuTronix]

cskamil,

Thanks for the information. Your B3 Stepping explains the majority of the problem. SpeedFan is using the same Tjunction Max 100c value as it does for the G0. All 65 nanometer Tjunction Max values remain undisclosed by Intel, but the B3 (which is a pair of E6600 B2's) is 85c to 90c, which you've already observed and confirmed with Real Temp.

Regardless, it wouldn't hurt to check your IHS and Heat Sinc with a razor blade before proceding with SpeedFan calibrations. Even if you don't need to lap, at least you'll have a visual on the flatness of both surfaces. Be sure to check several different geometric planes and positions accross the surfaces against a strong light.

After you reinstall your processor and Heat Sinc, since you've already completed Part 1 of the calibrations, go ahead with Part 2, then report Tcase and Tjunction Idle and Load results. I'm sure you'll find that your temperatures will be accurate, which will then allow you to proceed with your overclocking efforts.

Comp

 

[husky mctarflash]

I have nothing of value to contribute here, other than to tell you that Computronix is a recognized authority on temps. This is the tomshardware equivalent of posting a question about apartheid, and having Nelson Mandela post the answer.

 

[CompuTronix]

husky mctarflash,

You just made my day! That's the nicest compliment I've had in quite some time, and I thank you very much!

I work hard at this topic for everyone's benefit, and you might even say that I'm obsessed with Core 2 temperatures. Since I began writing about Core 2 temperatures here at Tom's nearly 2 years ago and compiling data for the Guide , it's been like peeling layers from "The Sacred Blue Onion" to expose the facts about temperatures. I can tell you that Intel's policy of nondisclosure makes it extremely difficult to reveal useful information.

I'm committed to conducting ongoing research and field testing for this topic, and to continue improving the Guide for enthusiasts and the overclocking community. Please let me know if you have any questions or suggestions.

Thanks again,

Comp

 

[cskamil]

Sorry for late reply,

Again thanks for your replies. I understood from your post that my CORE temperatures are seen wrong on speedfan because of just uncompleted calibrations. Am I right?
I checked the both surfaces. On the cooler surface, there are some scratches. Those scratces are outside the area that directly contact with the CPU surface. I scared to check the IHS with a blade :s.

Thanks for your help. I appreciate that

 

[CompuTronix]

OK, then use whatever type of straight edge that you're comfotable with.