Different core temps: CPUID vs SpeedFan

[wpcoe]

I'm curious why the core temps are reporting a 10° difference in CPUID Hardware Monitor vs SpeedFan.

The fans speeds report identically in both programs, as well as CPU temp. Just not the core temps. Any idea why, and more importantly, which one is more likely to be correct?

 

[aigomorla]

rule of thumb.

Go with higher.

You're always safer going with the higher.

If your still nervous get another program like core temp... or real temp, and compare those against what u have.

 

[Maximilian]

I asked this question before but i forgot the answer lol, i dont remember why but speedfan is wrong, the temps are 8-10*C higher than what speedfan says. Realtemp is what i use and itll likely give the same result as your higher temp in CPUID.

The way temp sensors measured things or worked or something changed between the pentium 4 era and core2/i7 era. Everyone used to use speedfan with older cpu's and i guess speedfan is still using the same old method of temp recognition. Something along those lines i think.

 

[Gillbot]

software readings are notoriously inaccurate anyway. Most accept that realtemp is the closer of all the programs if you take the time to adjust TJMAX to the proper value.

 

[ehume]

Make sure you're running the most recent version of sf. I remember reading that the reporting method was changed . . .

 

[wpcoe]

Thanks for the replies. I guess I'll go with the bottom line that software temp reports are basically unreliable, and just use them as a relative source of resource. i.e. I can tell if an OC has raised the temp, just not by exactly how much nor what the resultant actual temps are.

Masochist that I am, I also ran CoreTemp this morning and got a *third* set of temps:

 

[mb103051]

those speedfan temps are way too low.. unrealistic...the 30c temps are more believeable..

 

[ehume]

Thanks for the replies. I guess I'll go with the bottom line that software temp reports are basically unreliable, and just use them as a relative source of resource. i.e. I can tell if an OC has raised the temp, just not by exactly how much nor what the resultant actual temps are.

Masochist that I am, I also ran CoreTemp this morning and got a *third* set of temps

Try Real Temp. When you have your fourth set of temp reports, you can discuss them here.

 

[wpcoe]

My money is riding on CPUID HW & RealTemp, and they both report 44° and 46°.

CoreTemp uses TJ Max of 95° whereas RealTemp uses 100°.

Then I googled and found specs (at http://www.tomshardware.co.uk/intel-dts-specs,news-29460.html) which show the TJ Max for my e6750 chip is 80°:

Following that, I found a fascinating thread here on anadtech.com: http://forums.anandtech.com/showthread.php?t=218917 which, although sometimes way above my head, included some things like:

"Originally posted by: rge
Intel has stated that most cpus (probably all 65nm) have significantly raised DTS offsets to prevent throttling below tcase, per presentation at IDF, see slides 7 and 13. The chart on slide if drawn to scale could easily represent 15-20C offsets.

Intel has stated the formula for calculating temps is
diplayed temp = tj target-DTS + DTS offset

For E8400, DTS offset = 0 as measured by IR, so temp = Tj target -DTS

For E6850 or E6750 GO, DTS offset = ~20C (constant) measured by IR (Tcase is 95C on E6850 GO when DTS=0)
displayed temp = tj target (80) - DTS + 20C offset
same thing as using tj 100(effective after adding offset) - DTS"​

and

"Originally Posted by Idontcare
I emailed the author of CoreTemp and asked why he doesn't code the program to use the correct/published TJMax for G0's and he said he knows its wrong but he did not like the idea of his idle temps being reported as being at or below ambient."​

(PS: The reason my temps are higher than in previous posts is that I'm not using the air conditioning today. I live in Thailand, and today is less humid and cooler: only 30.5° in my room today. Previous posts using air con, room temp was about 24°.)

 

[ehume]

I'm loving your reportage.

 

[Ben90]

Just don't use temps. Go by Distance to TJmax and never worry about if its accurate again. (unless you use phase change or something stupid, but even then it will always be more accurate than going by *C)

 

[wpcoe]

Since there was no sarcasm smiley, I'll say thanks. ;-)

I spent several hours googling things trying to find an "official" Intel TJmax for the e6750. After about a half-dozen .pdf files on intel.com my brain hurt. Then I found the anandtech.com thread I mentioned above, and lots of it became clearer. I'm sticking with CPUID HW & RealTemp for now.

 

[wpcoe]

(Previous reply was to ehume. Ben90 posted while I was composing, so *this* is a reply to Ben90):

Thanks, I hadn't yet realized that the bottom line is to pay attention to "Distance to TJmax". I guess my brain still is in agony a little bit.