Which do i believe?

Agathon

Distinguished
Jun 25, 2008
46
0
18,530
I have a e8400 and Real temp reads 36c, hardware monitor reads 46c, and speed fan says 40c. What can i trust? I'm so confused. =(
 

Agathon

Distinguished
Jun 25, 2008
46
0
18,530
Ok, that reads 45c. Are you sure this is reliable? What makes it more so than the others? Can i trust my reading in the bios?
 

dagger

Splendid
Mar 23, 2008
5,624
0
25,780
RealTemp always underestimates. Coretemp tends to overestimate, but should be more accurate than RealTemp. Besides, it's better to be on the safe side and overestimate.
 

CompuTronix

Intel Master
Moderator
Agathon,

The temperatures in your first post are correct. On 45 nanometer processors, Core Temp reads 10c too high, and SpeedFan reads 5c too high, unless you calibrate it by using my Core 2 Quad and Duo Temperature Guide, the link to which is provided below in my signature.

Regarding dagger'sstatement about the accuracy of Real Temp, he is simply uninformed or misinformed, as are so many other users. I've read the Real Temp documentation, as well as the entire Real Temp thread over at Xtreme Systems, which currently has 69 pages containing 1725 posts. I've also been corresponding with the authors of Real Temp, SpeedFan and Core Temp over time.

As the author of the Temp Guide here at Tom's, I can vouch for Real Temp, as it's accuracy is based upon extensive research, real world testing and analysis. Read the Real Temp documentation - http://www.techpowerup.com/realtemp/docs.php - to understand what it does differently than other utilities, and why it's the most accurate core temperature monitoring utility abailable.

Comp :sol:
 
Ok here we go...

The readings from the SpeedFan calibration were reported were either 2 to 3C higher or lower as read by the IR thermometer. As for RealTemp the average temp readings were higher or lower 5C than the IR thermometer readings. If you don't believe me or Comp or the RealTemp developer just buy a IR thermometer and see.
 

Agathon

Distinguished
Jun 25, 2008
46
0
18,530
Wow, thanks for the helpful and informative responses I really appreciate it.

CompuTronix,

You said the temperatures in my first post were correct. Which temperature were you referring to? I'm assuming by the rest of your post you mean the Real Temp's Temperatures? One last thing, are the temps in my bios considered accurate?

Thanks


 

CompuTronix

Intel Master
Moderator
Shadow703793,

Thank you for the details, and of course, I believe you. As I own several IR devices, and have been extensively testing temperatures on many new rigs since I originally wrote the Guide in Feb `07, I can appreciate the difficulties in obtaining accurate measurements, as well as major and minor variables involved in calculating temperatures.

Agathon,

Real Temp is correct, however, I was referring to the relationship between the three utilities which is typical, since SpeedFan is 5c higher than Real Temp, and Core Temp (least accurate) is 5c higher than SpeedFan.

Tjunction Max:

Real Temp = 95c (correct)
SpeedFan = 100c (5c too high)
Core Temp = 105c (10c too high)

Get it? Also, never assume that BIOS (which is Tcase, or CPU temperature) is correct. CPU temperature, which is read by BIOS, almost always needs to be calibrated. Further, understand that BIOS has nothing to do with Core temperatures. Read my Temp Guide to understand the differences between CPU temperature (Tcase) and Core Temperatures (Tjunction), and how to calibrate SpeedFan to achieve accurate CPU temperature AND Core temperatures to within 3c.

Comp :sol:
 

dagger

Splendid
Mar 23, 2008
5,624
0
25,780
The problem is, the system does not care what you think the temperature is, or in other words, what the actual temperature is. It will throttle when it thinks thermal threshold is reached, regardless of what calibrated temperature you get in a software application. Core temp use tjmax 100, which means overestimating, but it's in line with the system's perceived temperature. If one day you see a temperature that's 15 degrees below, yet the system still throttles, you'll know why.

Basically, the truth doesn't matter, it's only what whoever's in power perceives as the truth. Sounds like the real world, heh? :na:
 

CompuTronix

Intel Master
Moderator
dagger,

I certainly meant no disrespect, but you're a sharp guy, and you've been here at Tom's a short while, so please let's see if we can get your facts straightened out for you. While you're correct that Core 2 processors will throttle and shut down at Intel's undisclosed factory calibrated Tj Max specifications, revealing the thruth about Core 2 temperatures most definately does matter. Part of my efforts in helping people here on Tom's Forums is devoted to assuring that the most accurate and up to date information is being provided concerning Core 2 temperatures.

Only since the release of Real Temp, have the authors of Core Temp and Everest recently adopted "calibrations", which is a direct knee-jerk reaction to the research, testing and analysis performed by the author of Real Temp. SpeedFan is the exception, since it's always provided the means to offset, or "calibrate" temperatures, however, some of it's default Tj Max values are also incorrect for certain processor variants. Since Core Temp and Everest simply assume the Tj Max specs from mobile processors, which are available from Intel, they don't translate very well for desktop processors. Agathon's questions focus on his 45 nanometer E8400. While the ever popular "Core Temp" uses a Tj Max value of 100c for certain 65 nanometer processor steppings, it also uses a Tj Max value of 105c for 45 nanometer processors, which just doesn't add up, especially when using Intel's formulas for calculations.

Either way, Core Temp has been proven wrong with regard to what I've just descibed above. Core Temp is, however, still accurate on B2 stepping 65 nanometer processors, which are Tj Max 85c, and concurs with Real Temp in that instance. Just read the Real Temp documentation - http://www.techpowerup.com/realtemp/docs.php - to see for yourself the problems that Real Temp solves concerning sensor linearity and accuracy, which the other popular utilities simply do not address, and instead mask. Additionally, the Real Temp thread over at Xtreme Systems - http://www.xtremesystems.org/forums/showthread.php?t=179044 - contains posts from a few extremely sharp and well informed members who have shared privileged inside information from Intel, which once again validates the accuracy of Real Temp from a completely different perspective.

Understand that I've made it my mission to continually and meticulously research this topic, which I've done for nearly 2 years. I've carefully studied hundreds of Intel papers, and I'm very serious about my endeavors here at Tom's, so I'm not inclined toward distributing misinformation fabricated from ambiguous assumptions. As I mentioned above, although you've been here at Tom's for just 4 months, we're not excactly strangers. You've seen and read many of my posts, as I have yours. As such, you're familiar with my style of writing, so you know that my answers and explanations are not vague one-liners, which is why my post count isn't higher. I take great care to provide very thorough and detailed explanations, so that everyone has a clear understanding of the questions, definitions and answers.

Although there was a time that I endorsed Core Temp in my Guide, I no longer recommend it to anyone due to my own research, testing and findings, which also coincides with the research that went into the development of Real Temp. If you're still not convinced of Real Temp's accuracy, then consider yet another perspective; since the techniques I developed in my Guide for calibrating SpeedFan don't use Tj Max values to achieve accuracy, then why, after SpeedFan calibrations have been completed, does the Core temperatures agree with Real Temp within a degree or two, which is what Shadow703793 and many others have confirmed? Check it out for yourself. Try my SpeedFan calibrations then compare your Core temperature results with Real Temp.

The bottom line is that Real Temp does not under-report Core temperatures. The higher-is-better-than-lower school of thought doesn't substitute for accuracy, and therefore doesn't apply. Just because Tjunction Max @ 95c has a higher thermal limit than Tcase Max @ 70c, it doesn't suggest that overtemp is OK. As explained in my Guide, Core temperatures which approach Tjunction Max are already far too hot for stabe everyday use. The root of the problem is that users are so brainwashed by Core temperatures, they don't understand the relationship between CPU temperature and Core temperatures, which according to Intel documents is specified at 5c, and function independently from different sensors. From my Guide:

-Tcase/Tjunction-
--70--/--75--75-- Hot
--65--/--70--70-- Warm
--60--/--65--65-- Safe
--25--/--30--30-- Cool

Tcase Max (CPU) is the operational thermal limit specified by Intel, NOT Tjunction Max (Core). Since Intel specifies 72c Tcase (CPU) for the E8400, this makes the corresponding limit 77c for the Cores (Tjunction), which is far below 95c. In order to calibrate and monitor CPU temperature AND Core temperatures so you can see the entire thermal big picture, SpeedFan is the best freeware monitoring utility available.

Regardless, Real Temp is the most accurate utility currently available for monitoring Core temperatures only. If you doubt my conclusions and want to debate this further, then log on to Xtreme Systems, and discuss your reservations with the author of Real Temp, Kevin Glynn, who's username is unclewebb. I've followed his postings on other websites long before he created Real Temp. I think he's quite intuitive and has excellent insights. I've been corresponding with him through emails and PM's, as well as postings on his thread. He's very interactive, so you'll find that he'll quickly answer your posts.

I hope this explanation helps you, as well as others, to reach a better understanding of this issue.

Comp :sol:
 

CompuTronix

Intel Master
Moderator
Shadow,

As much as I'd like to speculate within the limits of wishful thinking, there are several posibilities:

(A) Nothing will change.

(B) Core temperatures sensors will become even less accurate at Idle.

(C) Core temperature sensors will be improved, Tjunction Max values will be disclosed by Intel and the CPU temperature sensor (Tcase) will be discontinued.

(D) Core temperature sensor registers will be masked and function in the background for throttle and shutdown only, and the CPU temperature sensor (Tcase) will be improved with an "autocal" feature, or be upgraded to a new DTS designed and dedicated specifically for temperature monitoring.

4 cylinder engines don't have 5 temperature guages, and Cores don't have Intercoolers! A single temperature guage works just fine.

Comp :sol:


EDIT: Would you care to comment on the explanation for dagger?
 

charlyv

Distinguished
Jul 29, 2008
1
0
18,510
Hi everyone.
I'm new to this forum but have over the last year been occasionally looking in. So what's motivated me to join your "disreputable bunch" is this discussion about CPU core temperatures. I have an E8400 and in Realtemp the mean temperature idle is 11 celsius, under full load (Orthos, CPU and RAM testing) it is 35. This would seem perhaps slightly low at stock but this is at 4Ghz. Temperature with Coretemp and CPUz are 30 degrees higher. My question - how come?
System...
E8400
FSB - 445
Foxconn X38A
Tuniq Tower 120 cooler
Antec Nine Hundred case

Charles
 

unclewebb

Guest
Sep 11, 2007
247
1
18,865
Thanks CompuTronix for your solid backing of RealTemp. Coming from someone that has put as much time, effort and research into this subject that you have means a lot to me.

Just wanted to inform you and others that I finally got RealTemp 2.70 uploaded at TechPowerUp. There's been plenty of positive feedback already and users like the new high temperature alarm feature that you recommended to me.

http://www.techpowerup.com/realtemp

Stability is excellent so far even after 32+ hours of use. Keep spreading the word and I'll keep working on making RealTemp the best core temperature monitoring program out there!

Here's how things looked after 9 hours of having RealTemp and CoreTemp both minimized to the System Tray. CPU Usage Time speaks volumes about which one I'd rather have in my system tray. :)

corerealtemphx3.png


charlyv: Post a screen shot of CoreTemp / RealTemp and CPU-Z. A 30C difference is impossible unless you have made some major errors in your calibration of RealTemp.