New coretemp show Incorrect temp(I think)

[commodore]

The new CoreTemp readings for M0 steppings are correct, read this for more info: http://www.overclock.net/intel-cpus/263 ... ronix.html

The Tj max is not 85c for these series of procs, Tj max is the temp the CPU will shutdown at, let your CPU heat up to 85c and you will notice it will NOT shutdown, but will shutdown at 100c because the Tj max is 100c.

Just read the guide above for complete info, it will make sence once you read it, yes the whole thing. I am going to save a copy of CoreTemp 0.96 incase Tj max is reduced to 85c again which is incorrect...

Hi, I am the author of the first thread linked by you @ [H]ardforum.

After a lot of reading up on the Tjunction Max issue for M0 stepping C2D's i'm still not convinced that the Tj Max = 100'C is correct. Why? because in your example you are assuming that Core Temp is reporting accurately before-hand and then confirming the same with the throttling temp of ~ 95'C. This apart from the fact that after lapping the IHS and my Tuniq Tower (which i thought were making very poor contact) i'm getting only 3'C lower than before...

@2.2Ghz (stock 1.28 Vcore) -->Core Temp 0.96
IDLE: 44'C
LOAD: 66'C

@3.3 (1.464 Vcore) -->Core Temp 0.96
IDLE: 49'C
LOAD: 81'C

Core Temp 0.95.4 assumes a Tj Max of 85'C and reports ~ 15'C lower values consistently THROUGHOUT.

Now, doesn't Core Temp read the Delta to Tjunction Max value from the CPU and then calculate the Tjunction temp.? So basically its result is based on the Tjunction Max value.

where, Temperature = Tjunction Max - Delta (DTS value)

If you take a Tjunction Max of 85'C the throttling temp. is bet. ~ 80-85'C , above Tcase max of 73.3'C by a delta of ~10'C which is dead on for the M0 stepping!


..or so i think

EDIT: Check this link about Core Temp
http://www.overclockers.com/articles1378/

[cpmee]

where, Temperature = Tjunction Max - Delta (DTS value)


Thanks for posting the article link.


This is where I question the premise of the formula for all steppings. Has it remained the same ?

[The Coolest]

Calculating the temperature is indeed done like so:

Temp = Tj. Max. - DTS value.
So basically when you change the Tj. Max to a higher value, the temperature reported would be higher by the same delta.
That's why I believe the throttling tests don't prove much of anything as it could just as well be that the CPU started throttling at 80C as mentioned above by commodore.

[fgw]

Calculating the temperature is indeed done like so:

Temp = Tj. Max. - DTS value.
So basically when you change the Tj. Max to a higher value, the temperature reported would be higher by the same delta.
That's why I believe the throttling tests don't prove much of anything as it could just as well be that the CPU started throttling at 80C as mentioned above by commodore.

art, could not resist in jumping in here:

in the old days of e6300, DTS reported the delta to TCC activation threshold - as of intel documentations. has this changed for new core processors? is dts now reporting a delta to tjmax, or is this just a naming issue?
may be we are talking about the very same temperature level. some people call it tjmax and some might call it tcc activation threshold.

and a second question: in a reply to a post over on xs (http://www.xtremesystems.org/forums/sho ... tcount=323) you wrote:

One of my ideas was that maybe each core had it's own Tjunction, but looking at the dump tells me that they all have a Tjunction Max of 105C.

as to my knowledge, tcc activation threshold was not readable by software in the past. has this changed? tjmax is readable by software? can you tell by looking at msr registers what tjmax of a given processor is?

you have my pm if you want to elaborate on this one. there are still some open questions in the pocket - man i'm curious!

[commodore]

Thanks to 'The Coolest' for the clarification about the throttling issue..
Believe me i have read up a lot on this 'Tj Max' issue ..including Core Temp's authors posts to Intel techs on the Intel forums.. !

A guy @ [H]ardforum mentioned that you should give more importance to the Delta (DTS value) reported than the actual Tjunction temperature reading.

I tend to agree.. the Delta (DTS value) remains constant and is independent of many factors ..including the assumption of the Tj Max value. Its should ideally be 20-25'C...

[The Coolest]

Calculating the temperature is indeed done like so:

Temp = Tj. Max. - DTS value.
So basically when you change the Tj. Max to a higher value, the temperature reported would be higher by the same delta.
That's why I believe the throttling tests don't prove much of anything as it could just as well be that the CPU started throttling at 80C as mentioned above by commodore.

art, could not resist in jumping in here:

in the old days of e6300, DTS reported the delta to TCC activation threshold - as of intel documentations. has this changed for new core processors? is dts now reporting a delta to tjmax, or is this just a naming issue?
may be we are talking about the very same temperature level. some people call it tjmax and some might call it tcc activation threshold.

and a second question: in a reply to a post over on xs (http://www.xtremesystems.org/forums/sho ... tcount=323) you wrote:

One of my ideas was that maybe each core had it's own Tjunction, but looking at the dump tells me that they all have a Tjunction Max of 105C.

as to my knowledge, tcc activation threshold was not readable by software in the past. has this changed? tjmax is readable by software? can you tell by looking at msr registers what tjmax of a given processor is?

you have my pm if you want to elaborate on this one. there are still some open questions in the pocket - man i'm curious!

I think we're talking about the same thing, just different names.
It has always been Temp = Tj.Max - DTS value.

The Tj. Max is readable by software. Intel claims the value in the register is only valid for mobile CPUs. But as there is no better way to get the Tj.Max we have to use the same method for the desktop chips as well. That is why sometimes the readings do actually look wrong, and in these rare cases, they may well be wrong.

Thanks to 'The Coolest' for the clarification about the throttling issue..
Believe me i have read up a lot on this 'Tj Max' issue ..including Core Temp's authors posts to Intel techs on the Intel forums.. !

A guy @ [H]ardforum mentioned that you should give more importance to the Delta (DTS value) reported than the actual Tjunction temperature reading.

I tend to agree.. the Delta (DTS value) remains constant and is independent of many factors ..including the assumption of the Tj Max value. Its should ideally be 20-25'C...

I agree as well. The only reason I'm not doing this is that I believe it would cause a lot of confusion among users, specifically those who are less "techy". People want simplicity, Core Temp gives them that. I believe that most readings it provides are correct (or at least close enough), and in the cases where the readings are wrong there's not much that can be done, except maybe use the Delta to Tjunction option, which basically displays the actual value read from the CPU DTS.

[GigaByte]

Oh well.. back to inaccurate readings for M0 with 0.96.1. I got 0.96 for personal use

EDIT: Cool.. when underclocked to 1.2Ghz with 0.90 vcore my temp is now 2c below ambient. M0 users that use 0.96.1 use with caution...

[maf718]

Hi, as it happens I was doing some underclocking experiments yesterday with an M0 2140 and coretemp 0.96. (ie the 100°c tjmax version). With vcore of 0.9 and clock of 6 x150, cooling at maximum, the chip was idling at 17°c above ambient, as measured by coretemp. Full orthos load only added another 4°c, suggesting that the cooling was pretty effective.

This temperature seemed a bit too high to me but not 15°c too high. Could it be that the actual tjunction max for these chips is neither 85 or 100, but somewhere in between?

[imposter]

17c above ambient is good temps... nothing wrong with that. ss?

[The Coolest]

At such frequency and voltage 17C over ambient at idle is a lot.

[GigaByte]

My ambient is 22c... with correct readings I am 23c above ambient on idle.. thats 45c. At stock I idle at about 35c.. if I run at stock on 0.96.1 I idle at 20c with an ambient of 22c.. why is idling in the 40's hot to you? I seen idles in the 50's plenty of times.

How about adding a tj max selection, 85c or 100c?

[maf718]

Just double checked my temps at that underclock on an M0 2140.

Coretemp 0.96.1 = 1°c above ambient

Coretemp 0.96 = 16°c above ambient

Thats with ambient measured one inch in front of the cpu fan, all fans on full and the side of the case open.

[GigaByte]

Just double checked my temps at that underclock on an M0 2140.

Coretemp 0.96.1 = 1°c above ambient

Coretemp 0.96 = 16°c above ambient

Thats with ambient measured one inch in front of the cpu fan, all fans on full and the side of the case open.

Looks good but there is a problem, what about Tcase temp? Those are the core temps, Tcase is always lower than cores because of the delta. Example, its like putting a pot on the stove, lets say the burner is at 102c, is the pot at 102c? No, its probably at 100c or slightly lower because it is not PERFECTLY flat on the burner.. same thing with core temps and Tcase. The heat source is always the peak hottest point. Any lower is below ambient. Bang, conflict with 85c tj max.. and thats what iam trying to explain here.


This is directed to everyone here:

Source: http://www.overclockers.com/articles1378/index02.asp

"There are two more important values that have to be taken into account:


TCaseMax: This value stands for the maximum temperature at the dead-center at the top of the processor core or the IHS which will ensure 100% stability and safety.

Tjunction (or TjunctionMax): This value stands for the maximum temperature at the junction between the processor die and the PCB it sits on, this is usually much higher than the TCaseMax value.
It is important not to confuse these two when reading the DTS temperature. If the DTS registers a temperature higher than TCaseMax, that doesn't mean the CPU is in danger, as the TCaseMax stands for the highest external temperature the CPU can take. On the other hand, when your chip reaches the Tjunction then you should start to worry, as that is a temperature you wouldn't want your CPU to cross."

[maf718]

Hi Gigabyte, I crossed posts with you above, I wasn't meaning it to look like I was disagreeing.

In fact i agree. I'm led to believe that with sufficient underclock I should be able to get Tcase to very near ambient but Tjunction will always be around 10°c higher. If I use the cpu reading in speedfan to approximate Tcase I do indeed get about 1°c above ambient. Trouble is neither version of coretemp gives me anywhere near 10°c above this, but 0.96 is the closest.

[GigaByte]

Hi Gigabyte, I crossed posts with you above, I wasn't meaning it to look like I was disagreeing.

In fact i agree. I'm led to believe that with sufficient underclock I should be able to get Tcase to very near ambient but Tjunction will always be around 10°c higher. If I use the cpu reading in speedfan to approximate Tcase I do indeed get about 1°c above ambient. Trouble is neither version of coretemp gives me anywhere near 10°c above this, but 0.96 is the closest.

Thats good, your delta is not always 10c above, it is usually 10c +/-3c so 7-13c difference. Under load with excessive vcore and overclock the delta may exceed 16c which is normal. When heavily underclocked and volted you should see the exact 10c delta, or at least 1c from it. Having the Tcase 1c above ambient as you said is correct, at ambient is ok too even 0.1c above it is correct. Above ambient is above ambient. Because M0 has a 10c delta that puts the cores about 10c above the Tcase at all times. This is not possible with a tj max of 85c as the cores will be no more than +/-2c of the Tcase. If an offset is given to Tcase to lower it 10c that will put it way below ambient which is impossible. This is the root reason why M0 are 100c tj max. If you haven't gave a +15c offset to your cores in Speedfan I reccomend you do so, then you have proper temps. Go on and overclock! Thermal spec of 73.3c is Tcase not cores

EDIT: People seem to be going by the majority (no surprise there), majority aka almost all temp monitor software reads 85c tj max for everything except quads, but the majority is not always right as in this case. I tried to email Intel to ask what the tj max for M0 stepping (for a screenshot of proof) is and all I get is some automated crap.. I want an actual human response....

[maf718]

If anyone is interested here is a screenshot with my 2140 underclocked to 900mhz and undervolted to 0.9v, at idle, ambient 24°c:



And for reference the same chip overclocked to 3ghz, full load, same ambient:



As Gigabyte says, an 85°c tjmax is not possible for this chip. However, if I use the 100°c tjmax of coretemp 0.96 as in the screenshots, it seems to give a Tcase to Tjunction delta of 15c which always scales within +/-1°c whether I underclock, overclock or run stock at idle or load. The only way to get this delta to the accepted 10°c for M0 chips would be to assume the actual tjmax for my chip is 95°c. Is this likely or even possible?

[GigaByte]

No such thing as a 95c tj max CPU as of yet. Can you download Speedfan 4.34 beta 37 and post your temps again please? Your 15c delta is very strange for an M0 chip and my best guess is Speedfan 4.33 which is an old version isn't reading Tcase correctly.

[maf718]

Yeah, Speedfan 4.34 beta37 and 4.33 give the same readings, I have them running side by side atm.

I just wondered if the quoted 100°c Tjunction max was a nominal figure and could be varied somewhat batch to batch as Intel's engineers see fit? They don't seem very forthcoming on information in this area.

Otherwise Speedfan could be reading Tcase wrong by 5 degrees, would explain the delta, but I thought underclocking and undervolting the chip would result in a just above ambient idle Tcase, which Speedfan seemed to confirm.

Also would like to say thanks to The Coolest for developing coretemp in the first place, it really is appreciated.

[The Coolest]

In my opinion the best way to use this technology is how Intel intended it be used.
In Core Temp there's an option to "Show delta to Tj. Max". This shows the actual value of the DTS, and the amount of degrees in C° from the #PROCHOT.
The only reason this is not used by default is because I believe it will cause a lot more confusion amongst most users and the readings Core Temp and the likes currently provide are accurate enough for the most part.

[cpmee]

In the next version of Core Temp, could you include Tcase temp ? Tcase temp would be more comparable to the mobo thermistor readings of olden days IMO.

[The Coolest]

That would defeat the purpose of Core Temp, and it isn't trivial as it may seem. It means to add all the mobo sensors there are in the market.

[GigaByte]

In my opinion the best way to use this technology is how Intel intended it be used.
In Core Temp there's an option to "Show delta to Tj. Max". This shows the actual value of the DTS, and the amount of degrees in C° from the #PROCHOT.
The only reason this is not used by default is because I believe it will cause a lot more confusion amongst most users and the readings Core Temp and the likes currently provide are accurate enough for the most part.

I don't know how that shows "accurate" data, with 85 or 100c tj max the delta to tj max is different.. and I think you mean it shows the ammount of degrees in celsius from THERMTRIP# which is shut down temp, more commonly known as tj max. PROCHOT# is the thermal throttling signal, and you should still have a tiny delta to tj max left when throttling occurs if you have TM2 enabled (~5c from tj max).

[maf718]

In my opinion the best way to use this technology is how Intel intended it be used.
In Core Temp there's an option to "Show delta to Tj. Max". This shows the actual value of the DTS, and the amount of degrees in C° from the #PROCHOT.
The only reason this is not used by default is because I believe it will cause a lot more confusion amongst most users and the readings Core Temp and the likes currently provide are accurate enough for the most part.

Thanks again. This is the feature of coretemp that I rely on, and know it stays the same whatever the assigned value of tjmax. I guess I'm like most people in that I get some kind of intangible extra reassurance from seeing a temperature displayed in absolute terms rather than this way, and have been puzzled by the yo-yoing tjmax for M0 chips in the last few versions of coretemp. It certainly confuses people who don't know how coretemp works, if they think their core is 60°c they're not going to be as concerned about cooling as they were when it was displayed as 75°c.

Just for the record, are you going to revert to the 100°c tjmax (for M0 chips) in the next release of coretemp?

[GigaByte]

If he don't then stick with 0.96, if you can't find it I can put it on Filefront.

[maf718]

Agreed, still using 0.96 in preference to 0.96.1.