i7 ocs.. temps..

Caoder · 11-02-2009, 04:26 PM

ok sooo you have ~5 temps for the cpu under any cirumstance .. your cores and the overall cpu.. ie.. 22c cpu idle and 27-39 core temps..
is this showing that the integrated hs having poor contact with the actual chips :/? my load temps @ 4 ghz is 63-67 on the cores with a gtz... (not complaining about the temp) but i just wonder why there is such a large difference...

dejanh · 11-02-2009, 04:43 PM

Quote:

Originally Posted by Caoder

ok sooo you have ~5 temps for the cpu under any cirumstance .. your cores and the overall cpu.. ie.. 22c cpu idle and 27-39 core temps..
is this showing that the integrated hs having poor contact with the actual chips :/? my load temps @ 4 ghz is 63-67 on the cores with a gtz... (not complaining about the temp) but i just wonder why there is such a large difference...

What? If you are asking why the overall CPU temperature is so much lower, then I can tell you it is because that reading is not accurate. The only readings that are accurate are for four individual cores. The other reported temperature is at IHS and is not measuring the cores directly like the core temperature sensors are in fact...

Like right now I am LinX testing at 4.5GHz on an i7 D0 and the overall temperature is coming in as 60C-62C (not accurate) but the individual cores are coming in around 75C - 79C (accurate).

Caoder · 11-02-2009, 04:57 PM

i understand its not accurate.. but i was just wondering if this shows bad contact between ihs and cores or is it just something that just occurs...

shazza · 11-02-2009, 05:32 PM

The variation looks a little wide for your cores, but I do see as much as 7-9 degree differences in mine.

You can run the Real Temp sensor test to verify that the sensors are all responding as they should be.

Caoder · 11-02-2009, 05:57 PM

what's the sensor test's output suppose to tell me.. (just ran it).. had some background processes goin but here's the screen =x also attached is my hw mon showing the difference 24 c under cpu vs 38~ on cores...

Hondacity · 11-02-2009, 05:59 PM

Core temps are the directly reading the core temps, the ihs theps are lower coz it's in contact with the cold side - ie water block

shazza · 11-02-2009, 06:02 PM

I'm no expert, but my understanding is it tells you if your sensors are responding appropriately, or are out of whack. Also, if you get no/little movement, it means a sensor is not responding. Yours look good to me - better than mine in terms of consistency.

You can also use this to "calibrate" - adjusting Tjmax for different cores so they'll be more consistent - not something I've done, as I prefer to see the variance (and compare it to my original baselines if anything goes goofy).

Lot's of good info in the Real Temp thread by UncleWebb in the Intel section - he also is generally quite helpful about answering questions.

rge · 11-02-2009, 06:32 PM

CPU temp is read from a diode located in cpu between cores in die substrate. It is calibrated via the bios to read lower than its actual reading in die substrate, to approximate IHS temp at one particular load. So if calibrated correctly, it will read artificially low at idle since gradient varies from idle to load, but read reasonably accurate within several C of IHS temp once core temps are loaded, providing you are using same cooler it was calibrated for.

So if your cpu temp is calibrated correctly it should read at several C low at idle, correctly at 80% tjmax temps, then error again at anything above calibration point. This is because it is inaccurate to use a diode located in die, to try to guestimate IHS temps. (it is too expensive and impractical to put thermocouple in IHS). Not to mention most bios writers place low priority on calibrating cpu temps correctly, not to mention if its calibrated well on one setup and cooler, it will be off on another.

Since intel knew this was inaccurate from calibration reasons and inaccurate variance at hot spots, and was costing them efficiency (errors in temp measurement causing throttling to occur sooner or later than should), intel quit using these cpu diodes and instead rely on core temps in hot spots in die for throttling. CPU diodes will probably, eventually be phased out completely. Core temps are calibrated at the factory by intel, and are always the most accurate, especially at load.

At idle, stock settings, high end cooling, if core temps avg 39C, your IHS measured with thermocouple would be about 30-31C, or about 8C lower (was on 2 of mine anyways, and same on other 45nm cpus at low 2-5W idle, stock, speedstep enabled settings). Once you overclock, depending on cooling, may be 12+C difference between core temps and IHS temps.

Either way, I just ignore cpu temps, since it is not possible for them to be accurate except at one small calibration window. The diode itself could read accurately for the location where it sits, but since its calibrated to read a location where it does not sit, and given the gradient from die substrate to IHS varies with load, it can not be accurate except for one particular load and only for one cooler, the one it was calibrated for.

That being said yours it not that far from truth, may be few C too low, but by luck as much as calibration.

triggs75 · 11-02-2009, 06:45 PM

Your VREG temps are really high. I would be more worried about that then your CPU temps right now.

Caoder · 11-02-2009, 06:52 PM

vreg at temps cause of oc.. i have a 120 right above it drawing air out and one pointed right at it >_<... wcing them soon as i decide between koolance, bitspower or wait till enzo puts out their fc for the mobo ==.== i would go ek but there's no nickle plated ones.... and i dun think it'll be worth the cost to get em nickled...

triggs75 · 11-02-2009, 07:31 PM

Yeah I have the same board and with mine OC'd to only 3.69, I was running high on the VREG temps as well. I went with the Bitspower to cool them down.

Caoder · 11-02-2009, 08:28 PM

i'm tryin to go all nickle plated or clear w/ nickle on this rig soooo yea.. i can only wait for enzo or koolance or nickle plate the eks.. (bp only offers black pom)

Humminn55 · 11-03-2009, 07:16 AM

Core i7 and Core 2 processors have 2 different types of temperature sensors; a CPU case (not computer case) Thermal Diode centered under the Cores, and Digital Thermal Sensors located on each Core. The case Thermal Diode measures Tcase (Temperature case), which is CPU temperature, and the Digital Thermal Sensors measure Tjunction (Temperature junction), which is Core temperature. Since these sensors measure 2 distinct thermal levels, there is a 5c temperature difference between them, which is Tcase to Tjunction Gradient. Ci7's and C2Q's have 1 Tcase and 4 Tjunction sensors, while C2D's have 1 Tcase and 2 Tjunction sensors. Uncalibrated default temperatures are seldom accurate.

The first part of the spec, TCase, refers to a single measuring point on the Integrated Heat Spreader (IHS). Since a thermocouple is embedded in the IHS for Intel laboratory testing only, CPU temperature is instead measured using a Thermal Diode centered under the Cores. Maximum case temperature (Tcase Max) is determined by Spec#. The CPU case Thermal Diode is how Tcase is measured, and is the CPU temperature displayed in BIOS and SpeedFan.

Heat originates within the Cores, where Tjunction sensors are located on the hot spots of each Core. Most of the heat dissipates from the top of the Cores through the Integrated Heat Spreader and CPU cooler to air inside the computer. Some of the heat dissipates from the bottom of the Cores through the CPU case, which creates a 5c thermal Gradient toward the center of the substrate, where the Tcase sensor is located. This heat then dissipates through the socket and motherboard to air inside the computer.

Darkfire · 11-03-2009, 08:11 AM

Sorry for a slight thread hi-jack, but do you know if it is normal for an i7 to have a 5C difference between the cores themselves? Mine has two cores reading anything from 4C - 6C higher than the other two (Core 0 and 3 are higher)...

Conumdrum · 11-03-2009, 08:15 AM

Yes, no big deal to have a 5C diff. Not an issue at all.

Hondacity · 11-03-2009, 10:02 AM

Quote:

Originally Posted by Humminn55

Core i7 and Core 2 processors have 2 different types of temperature sensors; a CPU case (not computer case) Thermal Diode centered under the Cores, and Digital Thermal Sensors located on each Core. The case Thermal Diode measures Tcase (Temperature case), which is CPU temperature, and the Digital Thermal Sensors measure Tjunction (Temperature junction), which is Core temperature. Since these sensors measure 2 distinct thermal levels, there is a 5c temperature difference between them, which is Tcase to Tjunction Gradient. Ci7's and C2Q's have 1 Tcase and 4 Tjunction sensors, while C2D's have 1 Tcase and 2 Tjunction sensors. Uncalibrated default temperatures are seldom accurate.

The first part of the spec, TCase, refers to a single measuring point on the Integrated Heat Spreader (IHS). Since a thermocouple is embedded in the IHS for Intel laboratory testing only, CPU temperature is instead measured using a Thermal Diode centered under the Cores. Maximum case temperature (Tcase Max) is determined by Spec#. The CPU case Thermal Diode is how Tcase is measured, and is the CPU temperature displayed in BIOS and SpeedFan.

Heat originates within the Cores, where Tjunction sensors are located on the hot spots of each Core. Most of the heat dissipates from the top of the Cores through the Integrated Heat Spreader and CPU cooler to air inside the computer. Some of the heat dissipates from the bottom of the Cores through the CPU case, which creates a 5c thermal Gradient toward the center of the substrate, where the Tcase sensor is located. This heat then dissipates through the socket and motherboard to air inside the computer.

nice info...5c is not absolute i pressume..

when the cpu cycles...cores heat up...ihs is still cooler..

when i load my cpu ihs is around 45-55c ....cores are at 70-80(depends on my blocks)

rge · 11-03-2009, 12:39 PM

The 5C gradient he is referring to is from intel whitepaper on 65nm cpus, and is the 5C max gradient measured (by intel engineers at max load) from one of the die/core sensors to the cpu diode located in die substrate. He is not talking about a 5C gradient from die to IHS, but 5C gradient from die (hot spot) to cpu diode still in die substrate. No one knows for sure if that same 5C max gradient exists on load on 45nm from hot spots to central diode given different power and power density. Since i7s have a TDP 2x that and higher power density, one might think the gradient is larger, though with die shrink it is closer, so who knows for sure... its as good of a guess as one has though.

If you calibrate it to read correctly for its location as he suggests, it may then accurately read that location in die subtrate (which makes more sense) but will no longer approximate IHS temps...not that it does a good job of that anyways.

Edit: and if overclocking, the gradient will definitely be higher then 5C from hot spot to cpu diode in die substrate...the 5C was stock settings, at certain max loading programs on a 65W 65nm cpu. As power goes up, gradient increases...so if calibrating, has to be done at stock settings.