Schroedinger

Hmmm... TSK, do you know if that E2180 has a TIM rather than soldered IHS, and is it still running after this adventure? Because Gigabyte and I were wondering what happens if you aim an IR gun right at the core...

__________________
"The internet is text for people who can't read." -Andrew Eldritch

GigaByte

Care to ship me your camera?


Reguarding this thread I thought I share my mini thermal test I did this morning. I sort of messed up on it so I decided not to give it it's own thread as I will probably redo it soon. This test is done with the IHS on and AS5 as the TIM between the die and IHS. The fan for my CNPS9700 was unplugged with the heatsink still mounted on the CPU.

Everest = Tj max 85c
RealTemp = Tj max 95c
SpeedFan = Tj max 100c (look down in task bar near clock)







As I said I sort of "messed up" on this test, but some may still find it helpfull/interesting.

__________________

unclewebb

I just finished testing my new M0 E2160 and I have to admit that RealTemp is WRONG!

The measured TjMax was only 85C for this processor. It is not 95C. The Intel Thermal specs can't be used to try and determine TjMax.

I'll post an updated version of RealTemp later tonight for anyone with an M0 processor like the newer E4x00 series. I'm more convinced than ever that the L2 E4300 is TjMax=85C but I still need to buy one to test it properly. The E4300 generally need a (++) calibration to get the idle temps close to reality. Keep in mind that the Idle Calibration feature is only a band-aid solution for the DTS sensor that is simply not 100% accurate at reporting idle temperatures. It's a lot better solution than mis-adjusting the TjMax and having your temps displayed wrong from one end to the other.

My E2160 needed an Idle Calibration setting of (+) for the most accurate idle temps.



Not a bad little CPU for the money once you get the FSB set properly. 200 MHz must be a joke!

GigaByte

No don't... you make M0 steppings Tj max 85c and the core temps will be lower than Tcase. I don't know how many don't understand this, Tcase is NOT a heat source or a core, Tcase is the temp reading of the die surface. The heat of the entire CPU is coming from the cores that are INSIDE the die and the heat spreads throughout the die and into the IHS or heatsink if there is no IHS. You are basicly saying the temp of the air around the flame of a lighter is HOTTER than the flame itself.

E2xxx/E4xxx M0 steppings are 95c Tj max which you got correct in RT 2.11, L2 steppings of E2xxx/E4xxx are 85c Tj max. The Tcase max thermal spec of L2 and M0 stepping alone already signifies each one's Tj max. If you really insist on making M0 steppings have a Tj max of 85c which just don't fit period, show me how you are comming up with 85c for M0, because it just don't work.

__________________

Schroedinger

The temps will be lower than Tcase not because Tjmax is wrong, but because the DTS systematically overestimates degrees to throttle on Allendale cores if you are more than 30-35 degrees from Tjmax. Hence the calibration setting of "++".

The opposite is true of Wolfdale cores, according to Unclewebb's data; Wolfdale underestimates the variable "degrees to Tjmax" at low temperatures (IE - near ambient).

Now I haven't done any research on Allendale cores, but that's my understanding of how the data works out from a million of these threads I've read trying to figure out why my Wolfdale reads above 50 C underclocked and idling.

__________________
"The internet is text for people who can't read." -Andrew Eldritch

GigaByte

It has nothing to do with the calibration settings ++ or any of them. It is not just idle when cores are lower than Tcase its on load as well. The DTS has a set Tj max, why would it throttle when its more than 30-35c from Tj max? If that was true everyone's CPU would be throttling right now with the slightest overclock..

__________________

unclewebb

People are putting too much faith in the CPU reading or TCase reading that some boards are putting out. The core temp data from the Intel processors is far more repeatable and accurate. I know I won't convince everyone which is why you'll be able to adjust TjMax to whatever you want. I know the truth for my M0 and it is definitely not TjMax=95C.

GigaByte

Ok then, what is making you believe so? please.

__________________

unclewebb

The IR thermometer tells no lies. As I've said on XS, you don't have to take a processor all the way to TjMax to figure out what it is. With the M0 at 50C, the DTS shows 35, at 60C the DTS shows 25, etc., etc. I wish a few more people would buy one of these thermometers. As soon as you start testing, everything about the DTS sensors and where they're accurate and where they are not becomes pretty obvious.

You can argue that my M0 does not represent all M0 processors and that's certainly possible but I'm not going to believe that until you back it up with some actual readings.

Do my low MHz / low voltage test and try to get your processor running as cool as possible at idle. How many degrees below ambient are you when TjMax is assumed to be 85C? Idle readings that are a handful of degrees below ambient is normal. The DTS sensors are not 100% accurate at reporting low idle temps. You can boost TjMax to 95C but that will be wrong for your full load temps. RealTemp is the only program that lets you calibrate the temp sensors down low without forcing you to use the wrong TjMax.

GigaByte

Everest = Tj max 85c, Speedfan = Tj max 95c

Idle @ 3.5Ghz, 85c Tj max temps are totally wrong.. Speedfan is more like it



Load @ 3.5Ghz, again Tj max 85c is wrong and 95c is right.. Look at the Tcase and Tjunction as 85c Tj max..



Idle and underclocked to 1.2Ghz with 0.90v, ambient is 24c, with 85c Tj max cores are below ambient.. while Tcase is less than 1c above ambient which is correct. Above IS above.



Load and underclocked to 1.2Ghz with 0.90v, ambient is 24c, with 85c Tj max core 2 is STILL below ambient even at load. Both cores still below Tcase.



I don't know what else it takes, for some reason almost every processor is 85c Tj max to you. I also reccomend you do not allow the user to change Tj max in RealTemp 2.2, thats fighting the fire with a flamethrower. Its best to give users just one set of temps for their processor, even if they are incorrect. As you can see it is not just idle temps that are showing incorrectly with 85c Tj max, but load as well. M0 steppings are not Wolfdales and Yorkfields here, they are not known to be having any sort of this "mis-calibration" at idle that you say they have. Since every M0 stepping user gets the same results from the test I showed above you are going to have a hard time proving that every single processor is "mis-calibrated" at idle. Only a handfull of 45nms are having that problem.

RealTemp is a good program with great potiential, but you're going the wrong way by giving 85c Tj max to almost every processor out there, I will probably just keep 2.11 for private use as its got M0's correct

__________________