Solution and Trials with Core Temp and Conroe.

People seem to swear by Core Temp saying that this new fangled piece of software will give you 100% accurate results to what your CPU die is, and has killed the need for onboard motherboard sensors from the last 20 years. To this I say in theory this is a good idea but in practicality Core Temp is still a beta program with some room to change.

The reason that people seem to have temperature 10-15 degrees high is that the TCaseMax is set to 85 when I believe it should be 100. On my Core Duo it is 100. What Core temp is doing is taking a delta and subtracting that number from the TCaseMax to give you a temperature from each core. After numerous processor remounts and trying other things to get temperatures down, I have come to this conclusion. When running 2 different mobos with my E6700 I found that it was exactly 15 degrees higher than the internal temperature (infinity 975 and p5w dh) when using excel to map the data. This happened time and time again at given times from hardware monitors and Core Temp.

So today I purchased an e6300, volted it to 1.65 placed a digital thermometer between the IHS and the standard heatsink. Opened it up and let in run. I got into windows for about 3 minutes (processor is fried now good experiment). The digital thermometer read the temperature between 1-2 degrees different then that of the motherboard cpu reading again and again Core Temp was 15 too high. The other thing is on throttle watch, Core Temp read 100 degrees C before it started to throttle. Anyone else can try this experiment as well. The Proc will not throttle till 100 in Core temp. Thanks.

Very interesting....a lot of guys on this forum SWEAR that CoreTemp is the end-all, be-all in accuracy.

Great finding ! Thanks ! :clap:

Just don't forget too much "real" throttling will shortened yer cpu life. ;)

I also noticed that everest ultimate reported core temps the same as the core temp program. Could be it using the same calculations. It'b be nice it what the motherboard reports as the core temp was true. If so these run much cooler than opterons:)

thats why I bought a 6300. Its toasted and tbh I could give a :banana::banana::banana::banana:.

Interesting find Yoxxy. My E6400 was running at 72C according to Core Temp (stock volts, stock cooler @ 2.66GHz). If you subtract 15C, 57C would be nice...

There is a big difference between diode temperatures and those measured on the surface of the IHS, so I'm sorry to say your test is flawed.

I trust CoreTemp (and Everest Ultimate) more than anything else. Anyone who has a doubt I suggest try undervolting your CPU. Like this:

http://img243.imageshack.us/img243/1...rvolteddq7.png

I undervolted the E6300 to 1.225V (the lowest my motherboard allows) and got the above temps. Now, if you subtract 15C then you get something like 22~24C? Um.. That's just a degree lower or same as my ambient temp. And I don't believe that a working silicon with electricity supply is running at ambient temp.

Time and time again the CoreTemp became the most trust-worthy temp reading tool for me. (Since I can't directly shoot the thermo-gun @CPU die)

I don't know, have you tried the Intel thermal analysis tool? It reads similar to coretemp and everest, it just moves around more, but its much the same. ProbeII reads the same motherboard temps as everest (top one) and speedfan.

http://img169.imageshack.us/img169/3953/tempshp0.th.jpg

i had an interesting thing with my CPUs/mobos

i basically went from one motherboard DS3 to another DS3 and temps were completely different at roughly same ambient/same everything and many many remounts

i was seeing a good 10-15C higher temps by going from one to another using Coretemp................made me think about it a bit.....:confused: can't be right

Quote:

---

Originally Posted by G H Z

There is a big difference between diode temperatures and those measured on the surface of the IHS, so I'm sorry to say your test is flawed.

---

thats what i was gonna say...

not that core temp isnt wrong, but the test is flawed.

Agreed to a point that a lapped processor is as close as you can get to the die. It wouldn't make sense that all conroes are losing 10-15 degrees based on just the IHS. The other thing is where is TCaseMax coming from. When the similiar architecture of Merom is 100 and Conroe is 85? Also I have tried the Intel Thermal Analysis tool, since it is not made by Intel I guess more then likely it is a copy of Core Temp.

Quote:

---

Originally Posted by Yoxxy

Agreed to a point that a lapped processor is as close as you can get to the die. It wouldn't make sense that all conroes are losing 10-15 degrees based on just the IHS. The other thing is where is TCaseMax coming from. When the similiar architecture of Merom is 100 and Conroe is 85? Also I have tried the Intel Thermal Analysis tool, since it is not made by Intel I guess more then likely it is a copy of Core Temp.

---

well... TAT is older then coreTemp :stick:

Quote:

---

Originally Posted by Yoxxy

Agreed to a point that a lapped processor is as close as you can get to the die. It wouldn't make sense that all conroes are losing 10-15 degrees based on just the IHS.

---

you mean IHS, the core to IHS thermal material, and the increased temps due to a *HORRIBLE* mount by having the probe between IHS and heatsink...

again, i dont doubt that core temp is wrong, but the idea of a probe on the IHS proving core temp is wrong is crazy.

The mount was horrible on purpose. And Yes I mean as close as you can get between the IHS and Heatsink. That was also the point of driving the processor to 100 degrees before it throttled. This was also the second theory about it not throttling at 85 as if that was the TCaseMax it would. No one has disputed this point.

We know the solder on the core melts anywhere from 77C to 85C, you'd think it would throttle before then. Are you saying you measured 100C before throttling with the thermometer or with coretemp?

Core temp. Thermometer measured right around 83.

Quote:

---

Originally Posted by Yoxxy

Core temp. Thermometer measured right around 83.

---

Makes sense. You may be right.

Yoxxy, since you said the e6300 is for disposal :D, mind test this to get the "truth" out ? :)

The proper way with probe right in the center of IHS , btw Intel did that too for their CPU profiling and testing.

Example pic on D805 smithfield
http://www.overclockers.com/articles1312/8052.jpg
Courtesy of www.overclockers.com

If CoreTemp is suppoed to have TCaseMax at 100C instead of 85C, then your temperatures will be 15C HIGHER not LOWER!
e.g. -25C from CPU (value is always negative from the CPU)
100C-25C=75C
85C-25C=60C

I think CoreTemp may need a TCaseMax of 60C.
60C-25C = 35C - much more reasonable

Core Duos are designed for laptops, which run much hotter so CDs and Meroms TCaseMax = 100C.

Remember CoreTemp's accuracy in relation to the TCaseMax (whatever it really is) is purely based upon the accuracy of Intel's Digital Thermometer, which is IMHO more reliable than any other previous means of measuring temperature.

Please read the technical documents on intel's website and then post.

sory but i cant trust BIOS :



CPU: E6400 @ 3040mhz

Vcore: Default

Bloco : nexxxos xp silver

Temps : Coretemp >> 30c idle // 35 FULL ----SpeedFan & BIOS >> 15c idle //19c FULL



http://img358.imageshack.us/img358/2992/tempslj6.jpg



15c ??? :nono: :nono:



regards

Don't all of these programs read from the from the same sensor??? I use CPUCool for temp reading, but we all know that the true temp reading would be from direct contact with the doide. However, this test is quite informative and it's something to definately something to take into account. Specially sinc he sacrifried an E6300 in the process. Thanks.

mascaras: Speedfan 4.29 don't support 965P chipset yet. What you see there is the motherboard temp and an unknown one.

Quote:

---

Originally Posted by Vric

mascaras: Speedfan 4.29 don't support 965P chipset yet. What you see there is the motherboard temp and an unknown one.

---

not an unknown temp its cpu temp , I know what im saying ;)

http://img201.imageshack.us/img201/8349/2800nw8me4.jpg
:toast: :toast:

God, all this temp talk is making me dizzy, i still don't understand it all. lol

The overall temp should be below 60C ? right?
the 2 cores should be ??? below what? mine go up as high as 66C? when is the frying point? 75C?

Then all we really have to do is keep temps below 70c wile using core temp.

Quote:

---

Originally Posted by fornowagain

We know the solder on the core melts anywhere from 77C to 85C, you'd think it would throttle before then. Are you saying you measured 100C before throttling with the thermometer or with coretemp?

---

No solder does not melt at 77c to 85c.That won't even make water boil.For example lead has a melting point of 621 OF (327 OC) and tin melts at 450 OF (232 OC). Alloying 63% tin with 37% lead forms a eutectic that melts at 361 OF (183 OC). Note that this particular alloy's melting point is lower than either of the parent metals.http://engr.nmsu.edu/~etti/fall97/el...cs/solder.html

Rodman I hope your right about the 70c.Because I'm starting to think the same thing.Or at least close to that fugure.

Well..if you dont trust CoreTemp, why not just use Intels own. I doubt they dont have a clue about how to monitor THEIR own thermal diode :rolleyes:

And remember there is a huge difference between measuring the internals and outside of something with heat exsposure inside or outside. :fact:

Coretemp guesses a tiny bit it seems, due to it seems to try calculate temp a tiny bit in advance. But the difference at full and idle load is never more than 1-3C.

http://shintai.ambition.cz/pics/core1.jpg

TAT can be found here:
http://shintai.ambition.cz/files/tat.exe

Quote:

---

Originally Posted by Warpcore

No solder does not melt at 77c to 85c.That won't even make water boil.For example lead has a melting point of 621 OF (327 OC) and tin melts at 450 OF (232 OC). Alloying 63% tin with 37% lead forms a eutectic that melts at 361 OF (183 OC). Note that this particular alloy's melting point is lower than either of the parent metals.http://engr.nmsu.edu/~etti/fall97/el...cs/solder.html

---

LOL, oh dear, not all solders are made of lead and tin, its an LMP with Indium creates a much lower point for the liquidus phase (ooooh scary I even can even remember how to read phase diagrams and it may not be a binary alloy so check the ternary PTD) and people have removed IHS at that temperature 77C. Its totally liquid by 85C, plenty of these solders alloys melt in the 70 range, so be careful.

http://img245.imageshack.us/img245/820/soldermo4.jpg

Here's a pretty mad thread, I haven't seen one work after the surgery yet.

http://img226.imageshack.us/img226/7383/mvc007sqz3.jpg

BTW,water can boil at the temperature, I'll let you figure that one out.

Quote:

---

Originally Posted by Shintai

Well..if you dont trust CoreTemp, why not just use Intels own. I doubt they dont have a clue about how to monitor THEIR own thermal diode :rolleyes:

And remember there is a huge difference between measuring the internals and outside of something with heat exsposure inside or outside. :fact:

Coretemp guesses a tiny bit it seems, due to it seems to try calculate temp a tiny bit in advance. But the difference at full and idle load is never more than 1-3C.

---

Coretemp guesses in advance? Clairvoyant programs, I like that.

Take a closer look Coretemp says the TcaseMax is 85C, do you know what the Tcasemax is for an E6600?

TAT is old, it's not even designed for C2D's, its designed for mobile cpus with a much higher TcaseMax.

Quote:

---

Originally Posted by fornowagain

Coretemp guesses in advance? Clairvoyant programs, I like that.

---

:toast:

Ok I stand some what corrected.Thanks for the info.Water will boil with less than 100c if you lower the air pressure.Like ontop of a mountain lol.

Quote:

---

Originally Posted by fornowagain

LOL, oh dear, not all solders are made of lead and tin, its an LMP with Indium creates a much lower point for the liquidus phase (ooooh scary I even can even remember how to read phase diagrams and it may not be a binary alloy so check the ternary PTD) and people have removed IHS at that temperature 77C. Its totally liquid by 85C, plenty of these solders alloys melt in the 70 range, so be careful.

http://img245.imageshack.us/img245/820/soldermo4.jpg

Here's a pretty mad thread, I haven't seen one work after the surgery yet.

http://img226.imageshack.us/img226/7383/mvc007sqz3.jpg

BTW,water can boil at the temperature, I'll let you figure that one out.

---

i replaced one of the side panel fans with a Panaflo M1, and the two Core temps went down to 59-60C around 8 to 9C lower than before, Motherboard went from 33 to 29C now at 100percent load.

I guess i will just replace one more fan with a panaflo and the temps might go to 57C for both cores lol

Quote:

---

Originally Posted by fornowagain

Coretemp guesses in advance? Clairvoyant programs, I like that.

Take a closer look Coretemp says the TcaseMax is 85C, do you know what the Tcasemax is for an E6600?

TAT is old, it's not even designed for C2D's, its designed for mobile cpus with a much higher TcaseMax.

---

The TcaseMax for a E6600 is 60.1C. However TcaseMax is the outside temp if you check the spec PDFs.
http://processorfinder.intel.com/det...px?sSpec=SL9S8

And TAT is old? Its from april 27th 2006. And its designed for all CPUs that includes a digital thermal monitor. incl P4/P4D.

Also I doubt Intel suddenly developed a whole new thermal sensor just to do the same thing.

Well my BIOS reports temperatures about 15C lower than Core Temp, so I would agree that something reports temps 15C lower than normal, but not Core Temp heh.

Quote:

---

Originally Posted by fornowagain

Coretemp guesses in advance? Clairvoyant programs, I like that.

Take a closer look Coretemp says the TcaseMax is 85C, do you know what the Tcasemax is for an E6600?

TAT is old, it's not even designed for C2D's, its designed for mobile cpus with a much higher TcaseMax.

---

About Tcasemax, i think that the relative temperature reported by the integrated digital thermal sensor is not referenced to Tcasemax (that is an external temperature) but to Tjunctionmax, that is an internal temperature, much higher that Tcasemax; this is logical because the DTS gives an internal temperature. The Intel datasheet too gives recommendation on the method on how to measure Tcase, which means that the DTS temperature is for sure not Tcase.
So the temperature reported by TAT and coretemp are correct, the author of coretemp must only change the label Tcasemax to Tjunctionmax.

Quote:

---

Originally Posted by astaris

About Tcasemax, i think that the relative temperature reported by the integrated digital thermal sensor is not referenced to Tcasemax (that is an external temperature) but to Tjunctionmax, that is an internal temperature, much higher that Tcasemax; this is logical because the DTS gives an internal temperature. The Intel datasheet too gives recommendation on the method on how to measure Tcase, which means that the DTS temperature is for sure not Tcase.
So the temperature reported by TAT and coretemp are correct, the author of coretemp must only change the label Tcasemax to Tjunctionmax.

---

As you have, I've read the specs, "The Coolest" says Coretemp is derived from the TcaseMax MSR. Fair enough can't argue with that. But if it is, again "The Coolest" says that's the way Intel specifies it, I still don't see the relevance of junction temperature referenced against the maximum case temp TCM. So I agree and I've said it before, why isn't it referenced against the maximum junction temperature, which is of course higher. If you can find any documents with a hint of Tjunctionmax I'd like to see it.

If it "IS" TcaseMax like "The Coolest" says, then the derived temp is correct to spec, I just can't get my head around what it actually represents. Here's the thing, who gives a toss what the junction temperature is? Avoiding throttling is about all they're good for. All the HSF thermal dissipation calculations, system builders and not to mention me. Go on how much heat energy and casing temperature to deal with. If I need to spec a radiator and flow rate for an engine, do I care what the combustion temperature in the chamber gets to? no. I need the heat load and cylinder head jacket temp. So I'll stick to the Tc, like I have for the last 10 years. What would be useful to know is the delta offset from case to junction and then I can use core temps.

So I replaced my stock Intel cooler today with a Ultra-120 and a low speed panaflow fan, and the temps dropped only 5c under load with Coretemp, which means 70c instead of 75c...isn't that sweet?

For sure the DTS temperature is not referenced to Tcasemax. Let's explain the reason in detail. The DTS gives the delta between a reference temperature (let's name it T_star) and the current temperature of the core where DTS is integrated (let's name it T). So the output of the DTS is:
Delta_T=T_star -T
Coretemp takes Delta_T and read the temperature as:
T_Coretemp=TCasemax_tilde-Delta_T
where TCasemax_tilde is in fact not TCasemax but another reference temperature that Coretemp takes from a MSR of the CPU. Anyway from TCasemax_tilde (85 C) and T_Coretemp we can find Delta_T, that is the output of the DTS. Now in my case in idle state at 3.4 Ghz 1.4 V Coretemp gives 52 C. So Delta_T=33 C. If the DTS output would have been referenced to TCasemax of my processor, Tcase would have been 60.1 - 33 = 27.1 C, that is impossible because now at Naples where i live we have 30 C ambient temp....
So we can conclude that the DTS output is referenced to a temperature higher than Tcasemax, that is the maximum tempetaure that the core where the DTS is integrated can reach. Now it is not important the absolute value of this temperature, but how far from this value we are...so the only important measurement here is in fact Delta_T, and this value should be exposed by Coretemp.