Core Temp WRONG!?

[commandercup]

Thats such an old version of core temp. M0 chips ahve been corrected since then.

thank you so much lol!

now the temperatures are fine! they are almost exactly what was listed in everest/realtemp/speedfan

didn't even think to update core temp...

 

[Mussels]

use real temp. coretemp worked accurately on the original core 2 chips (65nm conroe, kentsfield and allendale) but its been inaccurate since then.

Realtemp has adjustments for the latest CPU's.

 

[dadi_oh]

use real temp. coretemp worked accurately on the original core 2 chips (65nm conroe, kentsfield and allendale) but its been inaccurate since then.

Realtemp has adjustments for the latest CPU's.

+1 on Realtemp. Coretemp was giving wrong readings on my E8200 and E8400 since Coretemp used the wrong Tj max assumption. The writer of Realtemp corrected the wrong assumption by actually measuring it.

 

[Mussels]

intel states numbers for the max temp, which is supposed to read like

Tjunction -40C (max temp) the problem is, without knowing what TJ is... you're screwed. the Realtemp maker simply went and tested naked core2 chips with an infrared thermo, to see the real TJ temps.

 

[unclewebb]

Here's the testing I did on my revision M0 E2160.

http://www.xtremesystems.org/forums/showpost.php?p=2844886&postcount=423

After seeing this I concluded that TjMax is 85C and not 100C like CoreTemp used to assume.

 

[Mussels]

Here's the testing I did on my revision M0 E2160.

http://www.xtremesystems.org/forums/showpost.php?p=2844886&postcount=423

After seeing this I concluded that TjMax is 85C and not 100C like CoreTemp used to assume.

thanks for that.

 

[theonetruewill]

Here's the testing I did on my revision M0 E2160.

http://www.xtremesystems.org/forums/showpost.php?p=2844886&postcount=423

After seeing this I concluded that TjMax is 85C and not 100C like CoreTemp used to assume.

OK you've finally convinced me to start using realtemp as my monitoring program. Your post as well as the neat addition "to-tray" feature (yeah I know we've been pissing you off there) clinched it.

People should probably have a read of this too if they're still decinding which is the correct one to use;

Unclewebb, I am the author of the Core 2 Quad and Duo Temperature Guide over at Tom's: http://www.tomshardware.com/forum/221745-29-core-quad-temperature-guide#

I first became interested in this topic of "apples and oranges thermal fruit salad" in November 06. I wrote the Temp Guide in early February 07 because users were very confused, and no one else was committed to taking on the task of cleaning up this incredible mess that Intel has so inconsiderately dumped on us, their customers. Since the topic sorely needed attention, my Temp Guide was immediately Sticky'd, and has been a work in progress ever since.

I can appreciate what an uphill battle this is for you, and I'd like to congratulate you on a truly outstanding accomplishment. I've been following your work for the past many weeks, and have of course read every word of the "Novel". I'm convinced that you've produced some "breakthrough" results, and I find your work both fascinating and thought provoking.

Over the past 15 months or so, I've tested dozens of Core 2 Processors, and have accumulated a great deal of data regarding thermal behavior. I'd like to offer my validation to your findings by stating that the "canned" values of the popular utilities many times just don't make sense, nor do they show general consistency among variants. Based upon my own research and testing, I'm certain that you're onto something very promising here.

When I began reading your thread, and the posts of a small group of very bright Forum members here at ExtremeSystems, I was inspired to write a major update to my Temp Guide. You may find it interesting to know that although I had included the immensely popular utility "Core Temp" in the Tools Section of the Guide since the early days, it is no longer present in the updated version. I have instead included a link to introduce Real Temp in the Troubleshooting Section of the Guide, hopefully with your blessings.

I have featured and continue to support SpeedFan in my Temp Guide, obviously because unlike other popular utilities, (except Real Temp), it can be Calibrated. If you haven't read my Guide, there is a Calibration procedure in Section 9, which approaches the problem of Core temperatures quite differently, by shifting the emphasis away from dealing with the elusive variable of Tjunction Max in it's entirety.

First, a standardized Test Setup (similar to yours) is used for maxiimum cooling and Auto frequency and Vcore. Second, Ambient is measured, Idle power dissipation and CPU cooler efficiency values are plugged into SpeedFan, which yields an Ambient to Tcase Delta accurate to within a degree or two. Lastly, Prime95 is used to provide a 10 minute Load, and + 5c Offsets for Tcase to Tjunction Delta are plugged into Speedfan, which yields Tjunction values that are typically just a few degrees above Real Temp, and a few degrees below Core Temp.

The +5c Tcase to Tjunction Delta value at Load is based on the following Intel document - Thermal measurement in the Intel Core Duo Processor - http://eda-publishing.imag.fr/spip/IMG/pdf/TMI23.pdf - which I consider to be the "Holy Grail". After studying hundreds of pages of Intel documents over the past year and a half, I'm convinced that this obscure 5 page paper is as close as Intel has come in a single document, to inadvertently giving away their closely guarded "secret". The significance of this paper was also pointed out by rge back on page 13, post #319 of the "Novel".

I have a few other ideas I'd like to share with you as to how I may be able to help you in our common quest for the truth about the Tjunction Max debate. I will contact you by PM to discuss these thoughts, and likewise, feel free to PM me. In the mean time, please consider me to be you staunch ally.

Once again, outstanding work, unclewebb!

CompuTronix

 

[unclewebb]

E8400 Test

Thanks for the positive feedback.

On some forums, some people think I've just pulled these numbers out of a hat or something but I did a lot of testing before writing RealTemp and I've done a lot of testing since. I haven't seen any real world testing from the competition.

When you see the testing I've done, like this one with my E8400, there isn't a lot left to argue about:

http://www.xtremesystems.org/forums/showpost.php?p=2883315&postcount=573

 

[Mussels]

Thanks for the positive feedback.

On some forums, some people think I've just pulled these numbers out of a hat or something but I did a lot of testing before writing RealTemp and I've done a lot of testing since. I haven't seen any real world testing from the competition.

When you see the testing I've done, like this one with my E8400, there isn't a lot left to argue about:

http://www.xtremesystems.org/forums/showpost.php?p=2883315&postcount=573

after reading the original thread about realtemp, i dont see what was left to argue - running a chip naked and seeing what temps it throttles at, is kinda... obvious? You did something that took a fair bit of effort and a mild amout of risk, and got a program with the most accurate temp readings so far. Its a far cry from the days of temp sensors in the CPU sockets...

 

[unclewebb]

Thanks. It turns out the risk is highly overestimated. My E8400 must have thought it was on a holiday when I introduced it to the test bed at 1600 MHz, 1.08 volts and less than 40C.

Soon after, the holiday was over and it was time to put it to work.

With all the early reports of degrading 45nm chips I was a little worried but it seems to have survived. It needs more voltage than the good 45nm chips but no complaints with how it runs.

 

[rampage]

Nice OC unclewebb

Also for no real reason i thought i would let you know that i have the same idle temp issue my E6750, i am underclocking my chip for a HTPC and am running it at 250 * 8 = 2000mhz @1.1v but your software is buy far the best at reading the chip at load or during normal use

 

[theonetruewill]

On a separate note -

ULTIMATE THREAD HIIJACK!

 

[unclewebb]

rampage: Read the RealTemp docs and try doing the calibration with your case open and compare your idle temps to your room temperature. If you have been idle for 5 or 10 minutes then there is usually about a 4C difference.

You should be able to find some Calibration factors for RealTemp that will get your idle temps much closer to reality. Even if they are not perfect, they will be much closer and will also improve the accuracy of your load temps up to about 60C. When I originally checked my E6400, it needed a +2 calibration. That sounds like a good starting point for you and then adjust it so both cores idle at the same temperature.

http://www.techpowerup.com/realtemp/docs.php

 

[sneekypeet]

What I am really curious about is how we go from temps on a E2200 to temps on a E8xxx, they have nothing to do with each other.

So we are just supposed to ignore that at load this thing reads in the 70*C sometimes?

 

[unclewebb]

I apologize for getting off topic. As the programmer of RealTemp, I thought people might be interested in the testing that is behind the program. There's not a lot of difference between the sensors on these chips. Very few are even close to accurate at idle and most are extremely accurate at full load.

If RealTemp says you are at 70C at full load then you are at 70C unless your sensors are damaged but few are. If a chip is running that hot, it's maximum overclock will be heat limited. It could be an issue with the IHS to core contact. That happens sometimes. My revision M0 E2160 was one of the coolest running processors I've had.

For far too long people have been taking whatever CoreTemp says as accurate and I just wanted to set the record straight.

I also wanted to show that you can test a processor this way without too much risk of damaging it as long as you use some common sense. I wish more users would step forward and do their own testing like this so more people can see that CoreTemp is WRONG sometimes which I thought was the topic.

I'll head back to XS and keep quiet here.

 

[theonetruewill]

Personally I am beginning to believe that my chip easily hitting above 70*C is just a bit odd since it's only on 1.45v and the cooler isn't that bad. The case has reasonable airflow and I think a temperature in the low 60's as real temp states is a bit more believable for my L2 revision E2140- even with the fairly large frequency increase.

 

[unclewebb]

The L2 E4300 is where CoreTemp first went wrong. He originally used TjMax=85C which is correct for that processor but users convinced him that the below ambient core temperatures that were being reported were wrong so they convinced him that TjMax was wrong. The only choices back then based on the documentation for mobile core processors was 85C or 100C. They said 85C was wrong so he chose 100C and L2 processors and many others have been reported wrong by CoreTemp and other programs ever since.

My M0 runs very cool and the L2 CPUs aren't much different. By testing a variety of processors, 45nm, 65nm, dual and quad core, early and late revisions you see a very clear pattern develop with these chips and its really not that complicated. The hard part is convincing users that the temps they've been seeing for the last year are wrong, and sometimes very wrong. As far as I know the latest version of CoreTemp 0.99 is still using TjMax=105C for the E8400. This combined with ignoring the trouble these sensors have at idle and my reported temps are off by 18C.

E2200 processors that are using TjMax=100C might have idle temperatures that are out by over 20C.