TSK's Crazy Theory on TJmax and the DTS - Page 7 - Overclock.net

**Retrospekt** · 03-14-08

Unclewebb: Were you able to push that cpu any further? Still interested in a trade for my L2 e4300?

Mikemoon07 · 03-14-08

dunno if this will help anything but this is what im seeing on my Q6600 B3 at 3006Mhz

**grunion** · 03-14-08

Quote:

Originally Posted by Mikemoon07

dunno if this will help anything but this is what im seeing on my Q6600 B3 at 3006Mhz

And that is wrong, use Coretemp.

**Retrospekt** · 03-14-08

Quote:

Originally Posted by Mikemoon07

dunno if this will help anything but this is what im seeing on my Q6600 B3 at 3006Mhz

Personally, I've just gotten used to trusting RealTemp. Unclewebb looks pretty confident in what he's doing, and the amount of testing he's done, I'm pretty sure he made an accurate program.

Mikemoon07 · 03-14-08

Those temps look really low for a quad at 3.0Ghz though. especially a B3, but this one does clock like G0's, must be a newer batch.

stumped · 03-14-08

I don't see why we are still relying on the Coretemp's Tjmax and RealTemp Tjmax.

Quote:

More than a few programs have been released over the last few years, each claiming to accurately report these DTS values in real-time. The truth is that none can be fully trusted as the Tjunction values utilized in these transformations may not always be correct. Moreover, Intel representatives have informed us that these as-of-yet unpublished Tjunction values may actually vary from model to model - sometimes even between different steppings - and that the temperature response curves may not be entirely accurate across the whole reporting range. Since all of today's monitoring programs have come to incorrectly assume that Tjunction values are a function of the processor family/stepping only, we have no choice but to call everything we thought we had come to know into question. Until Intel decides to publish these values on a per-model basis, the best these DTS readings can do for us is give a relative indication of each core's remaining thermal margin, whatever that may be.

If you guys really want the link to the thread in which it was posted, here it is

GigaByte · 03-14-08

Quote:

Originally Posted by Retrospekt

Personally, I've just gotten used to trusting RealTemp. Unclewebb looks pretty confident in what he's doing, and the amount of testing he's done, I'm pretty sure he made an accurate program.

As of 2.11 is was accurate for over half the processors, but now since hes throwing 85c Tj max at too many processors its loosing it's accuracy..

**Retrospekt** · 03-14-08

Quote:

Originally Posted by GigaByte

As of 2.11 is was accurate for over half the processors, but now since hes throwing 85c Tj max at too many processors its loosing it's accuracy..

He's using an IR gun, how can that be incorrect?

I know the L2's are 85c tjmax. 46c load at stock is ridiculous with CoreTemp.

GigaByte · 03-14-08

L2 and B2 are 85c Tj max. The rest are NOT as you can see it puts many processors BELOW ambient. An IR gun is not a fail proof tool, its most likely getting the temp of the IHS and at most the temp of the die if he removed any IHS's. Unless you know some magical way to create a very tiny IR gun and drill a hole in your die and stick it in. You are NOT going to get the temps of the the actuall cores that are in the die, you will be more like getting the Tcase which is always lower than the cores on all processors with any Tj max..

From the updated guide @ TomsHardware:

Quote:

Originally Posted by Computronix @ TomsHardware

Section 4: Thermal Flow

Heat originates within the Cores, where Tjunction sensors are located within the hot spots of each Core. Heat dissipates from the bottom of the Cores through the CPU Case, which creates a thermal gradient toward the center of the Die, where the Tcase sensor is located. Heat then dissipates through the socket and motherboard to air inside the computer case. Heat also dissipates from the top of the Cores through the Integrated Heat Spreader and CPU cooler to air inside the computer case. Safe and sustainable temperatures are determined by CPU cooling efficiency, computer case cooling efficiency, Ambient temperature, Vcore, clock speed and Load.

* Tjunction is always higher than Tcase.

* Tcase is always higher than Ambient.

Section 5: Findings

(A) Tcase is acquired on the CPU Die from the CPU Case Thermal Diode as an analog level, which is converted to a digital value by the super I/O chip on the motherboard. The digital value is Calibrated in BIOS and displayed by temperature software. BIOS flashes affect Tcase only.

(B) Tjunction is acquired within the Cores from Thermal Diodes as analog levels which are converted to digital values by the Digital Thermal Sensors (DTS) within each Core. The digital values are factory Calibrated and displayed by temperature software. BIOS flashes do not affect Tjunction.

(C) Tcase and Tjunction are both acquired from Thermal Diodes, while analog to digital (A to D) conversions are executed by different devices in separate locations. BIOS Calibrations, factory DTS Calibrations and temperature software can all be erroneous.

(D) Intel shows Maximum Case Temperature (Tcase Max) in the Processor Spec Finder. Intel does not provide documentation for Maximum Junction Temperature (Tjunction Max) on Core 2 desktop processors.

(E) Many popular software utilities assume Tjunction Max can be as high as 105c, which results in high Core temperatures and high Tcase to Tjunction Delta. Empirical data instead shows that C2Q and C2D variants range from Tjunction Max 85c to 95c, which results in lower Core temperatures and lower Tcase to Tjunction Delta.

(F) Although Intel does not provide documentation for Tjunction Max, 85c to 95c agrees with Intel papers which show that Tcase to Tjunction Delta peaks at 5c under Load.

Section 6: Scale

Safe and sustainable temperatures vary according to Spec#. The temperature Scales shown below illustrate the Delta between Idle and Load, and the typical 5c Delta between Tcase and Tjunction among C2Q / C2D variants. Although the Delta between Tcase and Tjunction is relatively consistent, temperatures do not always scale in a precisely linear manner with respect to one another, due to Variables such as Vcore, clock speed, Stepping and Load. Idle at very low Vcore and clock may cause Tcase to Tjunction Delta to indicate as low as 2c, while 100% Workload at very high Vcore and overclock may cause Tcase to Tjunction Delta to exceed 8c.

If temperatures increase beyond Hot Scale, then ~ 5c below Tjunction Max, Throttling is activated. The Digital Thermal Sensors (DTS) are used to trigger Intel`s TM1 and TM2 technologies for frequency, multiplier and Vcore Throttling within individual Cores. If Core temperatures increase further to Tjunction Max, then Shutdown occurs. Since Tcase indicates CPU Die temperature only, it is not used for Throttle or Shutdown activation, however, as Tcase Max will be exceeded before Tjunction Max is reached, Tcase Max is always the limiting thermal specification.

Use CPU-Z (see Section 8) to read processor information including the Revision field below the Stepping field, then choose a Scale below which matches the CPU being tested. Scales are ordered from highest to lowest Tcase Max.

Scale 1: Duo
E4x00: Tcase Max 73c, Stepping M0, TDP 65w
E2xx0: Tcase Max 73c, Stepping M0, TDP 65w
E8x00: Tcase Max 72c, Stepping C0, TDP 65w
E8x90: Tcase Max 72c, Stepping C0, TDP 65w
E6x50: Tcase Max 72c, Stepping G0, TDP 65w
E6540: Tcase Max 72c, Stepping G0, TDP 65w

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

Scale 2: Quad
Q9x50: Tcase Max 71c, Stepping C1, TDP 95w
Q9300: Tcase Max 71c, Stepping M1, TDP 95w
Q6x00: Tcase Max 71c, Stepping G0, TDP 95w

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

Scale 3: Quad
QX6x50: Tcase Max 65c, Stepping G0, TDP 130w
QX6800: Tcase Max 65c, Stepping G0, TDP 130w
QX6700: Tcase Max 65c, Stepping B3, TDP 130w
QX9650: Tcase Max 64c, Stepping C0, TDP 130w
QX9775: Tcase Max 63c, Stepping C0, TDP 150w

-Tcase/Tjunction-
--65--/--70--70--70--70-- Hot
--60--/--65--65--65--65-- Warm
--55--/--60--60--60--60-- Safe
--25--/--30--30--30--30-- Cool

Scale 4: Quad
Q6600: Tcase Max 62c, Stepping B3, TDP 105w

-Tcase/Tjunction-
--60--/--65--65--65--65-- Hot
--55--/--60--60--60--60-- Warm
--50--/--55--55--55--55-- Safe
--25--/--30--30--30--30-- Cool

Scale 5: Duo
E6x00: Tcase Max 61c, Stepping L2, TDP 65w
E4x00: Tcase Max 61c, Stepping L2, TDP 65w
E21x0: Tcase Max 61c, Stepping L2, TDP 65w
X6800: Tcase Max 60c, Stepping B2, TDP 75w
E6x00: Tcase Max 60c, Stepping B2, TDP 65w
E6x20: Tcase Max 60c, Stepping B2, TDP 65w

-Tcase/Tjunction-
--60--/--65--65-- Hot
--55--/--60--60-- Warm
--50--/--55--55-- Safe
--25--/--30--30-- Cool

Scale 6: Quad
QX6800: Tcase Max 55c, Stepping B3, TDP 130w

-Tcase/Tjunction-
--55--/--60--60--60--60-- Hot
--50--/--55--55--55--55-- Warm
--45--/--50--50--50--50-- Safe
--25--/--30--30--30--30-- Cool

Section 7: Parameters

(A) NO temperatures can be less than Ambient.

(B) Normal Ambient temperature is specified at 22c.

(C) Vcore Load should not exceed 1.4 volts on 45nm processors.

(D) Vcore Load should not exceed 1.5 volts on 65nm processors.

(E) All temperatures increase as Ambient, clock and Vcore increase.

(F) Tcase Idle is always higher than Ambient, even if less than 0.5c.

(G) Tjunction Idle average Core should typically be at least 2c higher than Tcase.

(H) Tcase to Tjunction Delta is typically 5c +/- 3c during Prime95 Small FFT`s.

(I) Tjunction Load should not exceed Hot Scale for the CPU being tested.

(J) Tcase Load should not exceed Hot Scale for the CPU being tested.

(K) Tjunction Idle and Load Results are average Core temperature.

(L) Idle to Load Delta may exceed 25c when overclocked.

unclewebb · 03-14-08

I'm afraid GigaByte that you are wrong about your M0 and all your posts only prove that to me.

First thing is you can't base any of your arguments on what the TCase temperature is being displayed as. That sensor is not calibrated or proven to be 100% accurate across its entire operating range. Unless you have some way to prove the calibration accuracy of that sensor then you can't use its data in any argument, either for or against. Just because a sensor gives you numbers that look believable does not make those readings proven accurate.

The most important data you have shown is that when you reduce the MHz and core voltage and let your processor idle, the reported core temperature based on TjMax=85C is a couple of degrees below ambient. That is normal behavior for idle temps based on the DTS. The DTS does not follow a linear line from idle to TjMax. They all diverge at low idle temps. Some start reading too high and some start reading too low. I first noticed this with my original E6400 Conroe core. It too can report idle temps below ambient. This is not a new problem with the introduction of the 45nm chips. It has been around since day 1 and it wasn't until about the time my program surfaced that Intel finally admitted to this. Head to Anandtech and read their most recent E8500 45nm story.

If you head over to XS you can also read the posts by rge who has done plenty of research to back up my findings that IR readings at idle on the IHS are almost exactly equal to the true core temperature of the CPU and within the +/- 1C accuracy of my Fluke. This is fact and has been proven in testing by Intel.

I released version 2.2 of RealTemp and I've set TjMax=85C for the M0 processors. You can keep using version 2.11 if you disagree or you can download version 2.2 and in the RealTemp.ini file you can set TjMax=2 which will boost TjMax up by 10C back to 95C. This might make you feel good but it is not accurate. Leaving TjMax=85C and using the (++) Idle Calibration will give you very accurate temps from ambient to beyond TjMax.

Will your temps be perfect at idle? Of course not. The DTS is simply not 100% accurate down low. My Idle Calibration will get you in the ball park though and your reported temps will be more accurate across the entire operating range than what any other core temperature software is providing.

Head to XS and read post#1 and you can also download from there.
http://www.xtremesystems.org/forums/...d.php?t=179044

		Overclock.net - Overclocking.net > Intel > Intel CPUs
TSK's Crazy Theory on TJmax and the DTS

03-14-08	#70 (permalink)
unclewebb New to Overclock.net Join Date: Mar 2008 Posts: 29 Rep: 8 Unique Rep: 8 Trader Rating: 0	I'm afraid GigaByte that you are wrong about your M0 and all your posts only prove that to me. First thing is you can't base any of your arguments on what the TCase temperature is being displayed as. That sensor is not calibrated or proven to be 100% accurate across its entire operating range. Unless you have some way to prove the calibration accuracy of that sensor then you can't use its data in any argument, either for or against. Just because a sensor gives you numbers that look believable does not make those readings proven accurate. The most important data you have shown is that when you reduce the MHz and core voltage and let your processor idle, the reported core temperature based on TjMax=85C is a couple of degrees below ambient. That is normal behavior for idle temps based on the DTS. The DTS does not follow a linear line from idle to TjMax. They all diverge at low idle temps. Some start reading too high and some start reading too low. I first noticed this with my original E6400 Conroe core. It too can report idle temps below ambient. This is not a new problem with the introduction of the 45nm chips. It has been around since day 1 and it wasn't until about the time my program surfaced that Intel finally admitted to this. Head to Anandtech and read their most recent E8500 45nm story. If you head over to XS you can also read the posts by rge who has done plenty of research to back up my findings that IR readings at idle on the IHS are almost exactly equal to the true core temperature of the CPU and within the +/- 1C accuracy of my Fluke. This is fact and has been proven in testing by Intel. I released version 2.2 of RealTemp and I've set TjMax=85C for the M0 processors. You can keep using version 2.11 if you disagree or you can download version 2.2 and in the RealTemp.ini file you can set TjMax=2 which will boost TjMax up by 10C back to 95C. This might make you feel good but it is not accurate. Leaving TjMax=85C and using the (++) Idle Calibration will give you very accurate temps from ambient to beyond TjMax. Will your temps be perfect at idle? Of course not. The DTS is simply not 100% accurate down low. My Idle Calibration will get you in the ball park though and your reported temps will be more accurate across the entire operating range than what any other core temperature software is providing. Head to XS and read post#1 and you can also download from there. http://www.xtremesystems.org/forums/...d.php?t=179044

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)