Question about odd temp readings (C2D)

[ShockwaveVT]

I finally got around to over clocking my E6300. Right now its running at 2.4Ghz (7x343) at Vcore 1.135 (BIOS set to 1.2625, lower than default). HSF is the stock Intel cooler, reinstalled with Arctic Silver 3 on an ASUS PK5-E mobo. Yes, AS3 is old, and a bit warmer than modern pastes, but its what I had laying around.

The thing that has me perplexed is not the temps overall, but rather that at both idle & load, the CPU temp is 4-6C higher than my core temps. Everything I've read says Tcase should be ~10C lower than Tjunction.
idle:
CPU (Tcase): 42C
Core0/1 (Tjunction): 37/37C

loaded with Prime95, multi-thread version:
CPU (Tcase): 53C
Core0/1 (Tjunction): 49/49C

The Intel thermal limit for this chip is 61C Tcase and I'd like to stay as far below that as possible for longevity. If I bump the vcore to increase my o/c, Tcase temps rise rapidly to the upper 50's.

My only thought is that I used too much AS3 when installing the CPU on my new motherboard, but I don't see how that would increase Tcase but not Tjunction. Any thoughts?

 

[Idontcare]

Unless I missed it in your post, you failed to mention the most critical aspect of your situation...i.e. how are you measuring your Tcase and Tjunction? Software, hardware? What software, etc.

 

[ShockwaveVT]

These temps are from Speedfan v4.33
Core Temp v0.96.1 displays the same Tjunction #'s as Speedfan for both cores.

oh ya and PECI is enabled on my motherboard

 

[Idontcare]

In this case neither is measuring Tcase. Tcase requires direct measurement of the IHS temperature.

Tjunction is inferred from the on-die DTS. Coretemp is well accepted as giving you the most meaningful temperature numbers. Albeit Coretemp reports the temperature as Tjunction which it is not. It is DTS.

See Appendix D, starts on page 75, of Intel's Thermal and Mechanical Design Guidelines spec document:
http://download.intel.com/desi.../designex/31559406.pdf

So as you can see, the initial concerns you expressed cannot be answered based on the data you have at your disposal as the data are actually neither Tjunction nor Tcase.

 

[ShockwaveVT]

...

Sorry if I confused you, I was trying to clarify that there is a difference between CPU & Core Temperatures. Many over clockers go by core temperatures only, but since Intel's thermal specification is based on Tcase, not Tjunction I am going by CPU temp.

Tcase is closely approximated by a CPU temperature diode built into chip, located beneath the IHS. This is what Speedfan reports as CPU temperature.

Tjunction is closely approxinated by the DTS embedded in each core. This is what Speedfan & Core Temp report as core temperature.

As I stated in the OP, its typical for reported CPU temp to be _lower_ than reported Core temp. In my system the opposite is occurring and I feel its might be limiting my overclocking overhead. I am looking for assistance in diagnosing possible causes, ramifications, and corrective actions.

 

[Tempered81]

@ Idontcare: when measuring Tcase, is the 72.3C on the Intel spec for an E8400/E3110 the Tcase maximium, or the Tjunction recommendation for core temperatures?

edit: I am curious because 73C tcase = 87C core temps...

 

[ShockwaveVT]

jaredpace, its my understanding that the thermal spec maximum is based on Tcase.

Tjunction Max (85C, 100C, 105C depending on processor) is the max core temperature and will induce throttling.


edit: at 87C your core temps are below the 100 (or is it 105) Tjunction Max for an E8400.

Its important to note that you can have core temps below Tjunction Max and still have a CPU/Tcase temp that is higher than the thermal spec. i.e. something like 90 core and 77 CPU.

 

[Tempered81]

tjunction max is 105C on e8400. so yes 87 is below 105 but that is not safe, correct?

Originally posted by: ShockwaveVT


Its important to note that you can have core temps below Tjunction Max and still have a CPU/Tcase temp that is higher than the thermal spec. i.e. something like 90 core and 77 CPU.

So the difference is not always 15C?

So it would be wiser to monitor Tcase while priming so as to keep it under the 72.3C spec. How would coretemp be useful when it isn't reporting Tcase temperature. Everest, on the other hand, will report Tcase temp.

 

[ShockwaveVT]

Those 77/90 #'s i posted were just an example, I wasn't going by any rule.

I guess in your 73/87 example you were assuming that core temps are always 15C higher than Tcase? If that is true then I guess Core Temp is useful. Its main attraction is that its a simple little program that doesn't wasted time checking other things like HDD temps.

But honestly Speedfan and Everest provide the same info that Core Temp does, and like you pointed out they provide CPU temp as well, and since Speedfan is showing me the same core #'s as Core Temp, I don't bother with Core Temp most of the time.