Real Temp - New temp program for Intel Core processors

[Degenerate83]

ok i'm confused by all off this, to much info to take in. running my e8200 o/clocked to 3.8ghz vcore 1.3625,
your real temp is saying cores are 31c 25c, while core temp is 41c 35c, 10degrees is alot. so do i need to adjust any settings like -- or ++
confused!!

[unclewebb]

Degenerate83: CoreTemp assumes a TjMax=105C for the E8x00 desktop processors which is based on what the mobile 45nm processors are rated at.

My testing of an E8400 with an IR gun showed that TjMax=95C for these processors.

Reported Temp = TjMax - DTS

DTS is the data coming directly from the on chip digital thermal sensors. Both programs read that data the same. Because both programs are assuming different TjMax values, the reported temps will be different by exactly 10C. Obviously I believe that RealTemp, which is based on measured data, is more accurate.

Chrysalis: The theory that the original B2 series has perfect accurate temps across the entire operating range is simply not true. I have an E6400 - B2 with the Conroe core. All programs agree that TjMax for the E6x00 series is 85C and my IR thermometer agrees but that results in temperatures at idle that are below ambient. Post#1 outlines this. The data output from the DTS sensors does not move linearly with changes in core temperature. RealTemp is the only program that recognizes this fact and tries to provide a way for users to calibrate their processor for more accurate idle and full load temperatures.

Other programs like CoreTemp are changing the TjMax value, usually higher, to try to cover up below ambient temperatures being reported. If you change the assumed TjMax, you will screw up the reported load temperatures.

rge: Thanks for your data. After measuring a M0 processor I was starting to wonder if the G0 series was TjMax=95C or not but your data confirms that it likely is. Hopefully another user can do some testing or else I'll have to go buy an E6750 or similar.

[Degenerate83]

cool!! so i'll be using this from now on then :P !!!!!

[Chrysalis]

Degenerate83: CoreTemp assumes a TjMax=105C for the E8x00 desktop processors which is based on what the mobile 45nm processors are rated at.

My testing of an E8400 with an IR gun showed that TjMax=95C for these processors.

Reported Temp = TjMax - DTS

DTS is the data coming directly from the on chip digital thermal sensors. Both programs read that data the same. Because both programs are assuming different TjMax values, the reported temps will be different by exactly 10C. Obviously I believe that RealTemp, which is based on measured data, is more accurate.

Chrysalis: The theory that the original B2 series has perfect accurate temps across the entire operating range is simply not true. I have an E6400 - B2 with the Conroe core. All programs agree that TjMax for the E6x00 series is 85C and my IR thermometer agrees but that results in temperatures at idle that are below ambient. Post#1 outlines this. The data output from the DTS sensors does not move linearly with changes in core temperature. RealTemp is the only program that recognizes this fact and tries to provide a way for users to calibrate their processor for more accurate idle and full load temperatures.

Other programs like CoreTemp are changing the TjMax value, usually higher, to try to cover up below ambient temperatures being reported. If you change the assumed TjMax, you will screw up the reported load temperatures.

rge: Thanks for your data. After measuring a M0 processor I was starting to wonder if the G0 series was TjMax=95C or not but your data confirms that it likely is. Hopefully another user can do some testing or else I'll have to go buy an E6750 or similar.

to be honest I find post #1 massively confusing. I dont think I should be fiddling with settings until I see a temp someone thinks 'looks ok'.

I went to the url that was pointed out in the thread here.
http://www.ocforums.com/showthread.php?t=543522

They stated this.

Tcase to Tjunction Delta is ~ 10c on B3 / G0 / M0 Stepping and ~ 15c on B2 / L2 Stepping.
and
Tjunction Max = 100c (B3, G0, L2, M0 Stepping)
Tjunction Max = 85c (B2 Stepping)

All utils are assuming 85 for my TjMax which if the above info is right is correct. If you are saying the info above is wrong then just simply tell me what the correct TjMax is for the 6420 B2 stepping. For your info my temps are not below ambient temperature it is quite cold here at the moment.

So if I am on the right TjMax why am I meant to hit the ++ button? My idle temps are maybe wrong i already understand this as well but as long as my load temps are correct thats what I see as important.

[rge]

to be honest I find post #1 massively confusing. I dont think I should be fiddling with settings until I see a temp someone thinks 'looks ok'.

I went to the url that was pointed out in the thread here.
http://www.ocforums.com/showthread.php?t=543522

The person that wrote that guide, has now updated his guide to reflect 5C max gradients....see here - http://www.tomshardware.com/forum/22...perature-guide

In his older link you posted, the gradients were assumed to exist based on a tjmax that was guessed at from mobile cpus, minus cpu temp guessed at in some cases (E8400) from a diode underneath the cpu socket, which is not even being cooled directly as tcase, and possibly in others by a diode between the cores and still not in contact with tcase. Still want to assume those gradients are accurate? I dont think the author does, see his updated link.

The only gradient you can be relatively certain of, is the minimal one at idle steady state without heatsink, because that is corroborated by intel documents. Unclewebb is using relatively accurate, repeatable measurements that reflect core temps by measuring temps in that manner, idle steady state, no heatsink. His tjmax should be accurate. If you want to know the gradient, use his tjmax for core temps, then place a thermocouple in IHS of cpu so IHS is still flat, and put a heatsink back on and compare the two. Barring that, I would not worry to much about gradients. If you want a guess at a gradient, you can calibrate cpu temp to ambient, but unless it is being measured at or near Tcase point (and it is not), its going to be a guess, some seem accurate guesses some not.

[unclewebb]

They stated this.

Tcase to Tjunction Delta is ~ 10c on B3 / G0 / M0 Stepping and ~ 15c on B2 / L2 Stepping.

That goes completely against what Intel testing has shown. There are times, depending on what software you are running and definitely at idle that the delta between TCase and the core temp approaches zero.

rge has been following this subject very closely and provides plenty of documentation. His post on the previous page of this thread is a good example.
http://www.xtremesystems.org/forums/...&postcount=384

If you are only concerned about load temps then leave the Idle correction at 0. Your B2 E6420 has a TjMax=85C so your load temps should be reported very accurately.

At very low idle temperatures, my B2 E6400 and many other B2, L2 and M0 processors will report temperatures slightly below ambient which is impossible for an air cooled processor. If a user notices this and is concerned about this then they can set either a (+) or (++) correction in RealTemp to boost their reported idle temperatures.

Anyone that wants to check for this can simply run their processor at 266x6 ~ 1600 MHz and drop your core voltage down to 1.10 volts. Turn up your CPU fan to max, open your case and try to get your CPU as cool as possible. A typical reading in this test is about 4C above your ambient temperature. If software is reporting temperatures below ambient during this test then it is wrong. It may become correct at full load but your idle temps are wrong and if you want a more accurate idle temp then use the RealTemp correction feature.

I don't think that is too hard to follow. Let me know anything that needs to be clearer.

[Ace-a-Rue]

Ace-a-Rue: yep, had subacromial decompression by scope 3 days ago (basically grind away some sharp bone from acromion so does not stab/tear/irritate my rotator cuff when I raise my arm above horizontal)

hmmm....your shoulder problem sounds just like mine....thanks!...get well very soon!

[lowtcell]

Just got a G0 Q6600. RealTemp 2.2 is reading cores 5c lower then CoreTemp 0.97 and Everest.

[Ace-a-Rue]

i noticed a 10C difference between real temp2.11 and coretemp, and that is due to, at least to my understanding, a 10C difference in TjMax.

EDIT:...still 10C difference between real temp2.2 and coretemp.

[Demo]

you have to change the offsets manually

[Ace-a-Rue]

if that post was for me, i do know there is a manual adjustment.

[unclewebb]

Yes, if you don't believe the TjMax that RealTemp is using is correct you can still use RealTemp with your own choice of TjMax. SpeedFan also provides this feature and I totally agree with it.

I think RealTemp is doing a pretty good job at its guesses at TjMax but there will always be users that disagree. If you are in that camp then open up the RealTemp.ini file and edit this line:

TjMax=0

Change the 0 value to either -3,-2,-1,1,2 or 3 and your TjMax will be adjusted for you by that value times 5. If you set a TjMax adjustment of -2, you will decrease the TjMax by 10 degrees. If you set this to TjMax=3, you will increase the TjMax that RealTemp uses by 15 degrees. I think this feature gives a user enough room to make whatever correction that they feel is appropriate.

This is a new feature just introduced with version 2.2 and is also explained near the end of post#1.

With no documentation from Intel to confirm TjMax, users need the opportunity to adjust this value as more information becomes available through my real world IR thermometer testing as well as their own. That is what RealTemp is based on. Users with their own IR thermometers and their own testing methods are now free to set up RealTemp however they like with idle adjustments at the low end or TjMax adjustments to cover the high end.

RealTemp is the only available software that accurately reports the core temperature of my 3 Core based processors from idle to TjMax and beyond.

lowtcell: RealTemp assumes your TjMax=95C while other programs are assuming TjMax=100C. That is where the 5C difference comes in.

[Demo]

@uncleweb

Any feedback on this

[Ace-a-Rue]

Yes, if you don't believe the TjMax that RealTemp is using is correct you can still use RealTemp with your own choice of TjMax. SpeedFan also provides this feature and I totally agree with it.

I think RealTemp is doing a pretty good job at its guesses at TjMax but there will always be users that disagree. If you are in that camp then open up the RealTemp.ini file and edit this line:

TjMax=0

Change the 0 value to either -3,-2,-1,1,2 or 3 and your TjMax will be adjusted for you by that value times 5. If you set a TjMax adjustment of -2, you will decrease the TjMax by 10 degrees. If you set this to TjMax=3, you will increase the TjMax that RealTemp uses by 15 degrees. I think this feature gives a user enough room to make whatever correction that they feel is appropriate.

This is a new feature just introduced with version 2.2 and is also explained near the end of post#1.

With no documentation from Intel to confirm TjMax, users need the opportunity to adjust this value as more information becomes available through my real world IR thermometer testing as well as their own. That is what RealTemp is based on. Users with their own IR thermometers and their own testing methods are now free to set up RealTemp however they like with idle adjustments at the low end or TjMax adjustments to cover the high end.

RealTemp is the only available software that accurately reports the core temperature of my 3 Core based processors from idle to TjMax and beyond.

lowtcell: RealTemp assumes your TjMax=95C while other programs are assuming TjMax=100C. That is where the 5C difference comes in.

actually, unclewebb, coretemp tells me that my QX9650 has a TjMax of 105C...that stays true up to coretemp 0.97...not a big deal except the author wrote his program to display 105C for 9650's.

i am kind of with you that the 45nm cpu's TjMax might be around 95C.

i read from someone who said, at least from what i recall, that we can add either 20C or 25C to Tcase and that should be very close to TjMax...i.e.,my Tcase is 64.5; adding 20C would put my TjMax at 84.5C, or, 89.5C if i added 25C...any thoughts on that supposition?

[lowtcell]

lowtcell: RealTemp assumes your TjMax=95C while other programs are assuming TjMax=100C. That is where the 5C difference comes in.

Yeah, I got that much. I'm just not sure which one is the correct one to to by. I hate assuming things.

[jason4207]

Just got a G0 Q6600. RealTemp 2.2 is reading cores 5c lower then CoreTemp 0.97 and Everest.

i noticed a 10C difference between real temp2.11 and coretemp, and that is due to, at least to my understanding, a 10C difference in TjMax.

EDIT:...still 10C difference between real temp2.2 and coretemp.

You guys are comparing a Q6600 (65nm) to a QX9650 (45nm). Q6600 was reporting 100*C Tjmax in most programs, but reports 95*C Tjmax in RT...5*C difference. QX9650 was reporting 105*C Tjmax in most programs, and reports 95*C in RT...10*C difference.

Until Intel releases more info we have to make assumptions and/or do more of our own testing.

Or you can just not worry about it so much:

As long as your processor is not throttling, it is running within the Intel specs. My one great theory still holds true. As long as you are Prime / Orthos stable and you are not throttling then you don't have to worry too much about temperatures. You will lose long term stability if you are pushing your processor too hard.

[jason4207]

Made another comparison with CT 0.97 and RT 2.1

Idle temps are detected with 1c difference between programs (RT showing 1c lower compared to CT) by using a ++ calibration on the idle temps.

But during a certain amount of heat generated the CT readings pull ahead by 2~4c difference from RT ( again RT reads lower with ++ calibration )

I tried booting my system to 266x6 at 1.275(lowest VID possible on my Q6600 G0) resulting in a +1/-1 reading on CT & RT. ( on 15mins of Prime95 v25.6 SmallFTT )

Returning to my OC settings of 400*8 @ 1.335v ( bios volts ) I noticed that RT's reading is comparable to CT upto 49c only when both programs start to indicate any value of 50c, CT seems to start pulling ahead upto 4c difference from RT.

Basing on my unreliable room thermometer here, im getting 31c for my ambient. Using RT with 0 calibration shows my lowest core at 29c, + 32 & ++35



Loaded



Idle


Since RT detects the TjMax of a Q6600 at 95c w/c a lot of people believe the TjMax is 95c instead of 100c or 105c i would assume that the 0 calibration reading on RT is the correct temp since im getting the correct temps when i deduct the delta to TjMax with the current TjMax detected

No, unclewebb is saying the temps do not scale linearly as you get further from Tjmax, so the "Tjmax-DTS=currentTemp" formula doesn't hold true as you approach idle temps. That is what the correction is for. If your room thermometer is accurate then your correction factor of ++ puts you at 4*C above ambient which is the suggested value. Since you are unsure how reliable your room thermometer is you cannot make any assumptions. You'll just have to wait until you can get a room thermometer you can trust more.

[Demo]

@jason4207

If you take another look at my pictures you can see both programs at the same IDLE temp, and both programs have the same 1:1 reading up until 49c, when the temp enters 50c the gap starts to increase upto 4c, but unclewebb has PMed me about it.

[jason4207]

Yeah, but you're using the ++ correction factor. The reason 100*C Tjmax was originally selected by most programmers is b/c of idle temp readings relative to ambient. So CT may be indeed be correct at idle, but at load it would be off by ~5*C b/c of an incorrect Tjmax. RT should be correct at load, but needs the correction factor to be accurate at idle. Again it goes back to the theory that DTS doesn't scale linearly.

[leffe]

Seems to work well with vista64. Shows temps within 1C compared to U-Guru.

[nike]

great, finally a programm which can let me measure the temp of my E8400. coretemp stucks with the temp and also everest and smartguardian are way off, 28c idle and 37c loaded with a TRUE seems like ok(no damn idea of load temps, coming from an opty 170 and dfi nf4 sli dr) heres a screen:



greetz

edit|| also the 2nd coresensor seems to be borked just as mentioned at the 1st post. it shows a 0 at the 2nd sensor so the thing is stuck and cant provide accurate temps, so at least 1sensor works :P

[KTE]

It seems too many people are getting confused by what is happening or being discussed here and thus accepting any low temperatures as accurate as they did for CoreTemp before, as they did for Motherboard Monitor once.. as they did for Next Sensors when it was released. Too many are going to use this program wrongly to try and get their lowest temp readouts or whatever suits them whilst not testing anything practically. If you do not test accurately, you have no way to know your temperatures. A CPU TDP is given by testing the hottest chip in each SKU bin at max lab. loads, many chips in that SKU will have much less TDP than the rated. The value is just a nominal generalization band. TDP decides your CPU heat, as with any IC and thus your cooling. 50W on any other IC being dissipated is the same as 50W on a Wolfdale, the temperature will be the same, as the heat energy released in joules per second is the same.

Let me make it clearer hopefully:
-The Tcontrol + offset is not a straight line in relationship to the actual core or IHS temperature.
-Internal core temperature is not the same as CPU case temperature under load, ever on modern CPUs.
-CPU case temperature calibration in software needed at 0C will not be the same calibration needed for the correct IHS temperature at 30C and neither of which will be the calibration needed to find the IHS temperature at 60C. In the same way, you'll have to calibrate it separately for each CoreTemp reported temp for even the slightest accuracy or you'll create a puddle.
-The reported values from any software have no way to be verified ascertain unless you have an accurate methodology, thermocouple, across the measured materials and temperatures being tested, to test shutdown temperature, throttle temperature, ambient case temperature, ambient heatsink temperature and idling temperature at the lowest MHz+volts, lowest MHz+mediocre volts and lowest MHz+high volts. This is minimum to test reliability of your testing, as well as the temperature being taken from within the core IHS. Your data would have to correlate to that of other known IC temperatures at given TDPs.
-It is very clear max core temperatures, max shutdown temperatures and max throttling temperatures have never been the same on any last 3 years desktop core SKU. Own a few same SKU processors to test and find out yourself. So ones max shutdown/throttle temp findings are quite irrelevant to be applied to anothers, regardless of CPU MFG PR. Each core has a separate absolute limit.
-So, in order to trip the TCC and/or have PROCHOT# asserted, you'll need to do some experimentation.
-You do not know whether your core TCC is working until you've pushed it to throttle/shutdown/fried. A retail E8400 so far returned did not have the TCC working, it died when pushed. Penryn lineup is known to have fudged DTS throughout it's lineup straight from the first non-ES samples, this cannot be mitigated by knowing your Tjunction_max either, as the DTS are wrong in their calibration in themselves and thus their feedbck you find your temp through. Some, very rare I've seen working. We had 152 deliveries of Wolfdale at our work which my boss rejected because of this and Intel has not supplied us the new batches they promised 6 weeks ago which they say have this problem fixed. The problem is corrected in C1 step with desktop and mobile CPUs AFAIK. Core 2 has perfectly accurate DTS feedbacks on the majority though.
-With faulty DTS, you can calibrate to add -15C to temps you see as 115C in software believing you're still only at 100C because TJmax has not yet been reached while you're burning the core at 115C or even 125C actual because your TCC is faulty and never kicked in. You're damaging your hardware past the limit it can withstand. Hence stay well below the limit compared to what these software show you unless you can ascertain a working TCC.
-1.4V TRUE air 20C heatsink ambient on a 95W TDP CPU core at any MHz will never ever be idling below 40C core temp. Ask Intel if you don't know this. If it does show you this, you have problems with your reported temps being inaccurate.
-There is a minimum ~10C delta between the hottest part of a core to the coldest, Intel, MB MFGs and AMD will confirm this, let alone the hot spot from the surface at load which can be quite a bit. I've attained measurements many times before this with more accurate ways than a IR therm., using the method Intel and industries use, N/T/K type thermocouples calibrated at fixed temperatures from 10-125C, fixed lab. conditions. The surface was always cooler than the actual internal chip temperature being reported by the DTS, esp. under load. That's why Intel mentions in their procedure for case measurements, that if you do not insert the thermocouple inside the groove as instructed, the temperature feedback will not be accurate for the Tcase measurement, not the DTS measurement. If you're going to level the Tcase surface temp. with the DTS temp. you will error, as that will only give you the Tcase_max if done correctly, a low temperature. You can measure your Tcase by seeing when PROCHOT# is asserted. If it's 95C, the Tcase_max will be what Intel has specified on their site for each processor, when Tjunction_max internally is reached. So for mine, Tcase_max is 71C and at that period, Tjunction_max will be 100C. 29C delta between the two, it is not much at low TDPs but increases with higher heat emission.
-Tcase_max is 71C for my quad at 95W TDP. Tcase_max exists specified dependent on chip TDP. That's because there is a large delta between Tcase and Tdie under load: for Q6600 G0 71C Tcase would equate to the throttling point internally, which is your individually calibrated Tjunction_max. Both would be reached at the same time. The Tcase_max is the maximum the core can withstand before damage as is the Tjunction_max, they are just measurements taken at a different place, one internally at the hottest part of the core and one externally embedded within the IHS.
-AMD RS780 idles 1.5W TDP, that's power dissipated as heat, and even with a heatsink, it is still idling 3C above ambient. Most of the lowest volts/MHz I've seen are still above 10W TDP, which won't make the chip idle less than +4C ambient at minimum on the hottest core. RD790 is a 10W TDP maximum lab. load IC and 8W TDP maximum software load and with a heatsink it still idles 6C above ambient temperatures on air, that should tell you what to expect. No chip or cooling can discern anything more than heat, all heat energy emitted is in TDP (Watts) units. The chip surface and its cooling only recognizes TDP and deals with it. TDP is heat energy being dissipated as Joules, 50W TDP on any chip is the same thing for the cooling. You can work out TDP after you've oc'd/uc'd and approximate your temps well.
-45nm Penryn is very susceptible to EM, TZDB, Qbd and TBBD (damage/degrade/breakdown with stress, current and voltage), it has a maximum safe voltage rating at around 1.357V published inside of the Intel guides (I don't remember exactly, but very close to that 1.35V figure). This is due to the material choice and fabrication node, they do not tolerate voltage much and I have not seen the SEM cross section of Penryn die, but there would have to be a lot of correction done even in order to allow over 1.2V without causing large damage to the cores. You will definitely damage the core with higher volts than this rating, how long before it degrades, not long, how long before it dies, no idea.

graysky: excellent additions you asked for, makes the program worthwhile now thanks to unclewebb.

I have a G0 Q6600 here, I don't use the + and ++ calibrations which are needed after you've found your Tjunction_max and you're idling below what it's supposed to be. FWIW, I've found my Tjunction_max long ago. I've already said each chip differs, as it does, and all CPU MFG's confirm this. Furthermore the offset value differs for each too as the gradient increases. For instance, my Q6600 G0 Tjmax is not 95C. The DTS are correctly working on mine, as when Tjunction reports 0C through the DTS MSR, PROCHOT# becomes asserted and the core starts throttling. DTS MSR reads 0 when Tjunction_max is reached. THERMTRIP# is supposed to assert 20-25C after PROCHOT# assertion temp. is reached. THERMTRIP# exists on all CPUs, as a fail safe and accurately calibrated method (rarely one does slip through without properly calibrated). Intel statements are readily available on this too. So say, my Tjunction_max was 95C, >115-125C internal (DTS) core temps. will make the system shutdown, all clocks halted automatically. This you can't stop and mostly you won't be able to pick up. Most of the time, the DTS temperature jumps very quickly after PROCHOT# assertion if not cooled, not in 1-3C but in 5-10C in a ms. When I tested mine, I did so in quite a few ways =>

First Idle without CE1, daily setting [Fig.1]:


I usually will run CE1, so it'll idle lower. Coretemp/EVEREST/HWMonitor assume a 100C Tjunction_max, and that's my idle voltage at maximum fanspeed, with 23C ambient socket temp.

Idle totally stock, no CE1, same conditions but lowest fanspeed [Fig.2]:


Already, as you can see, 95C Tjunction_max default on Real Temp is giving errorsome idle temperatures. Two cores are reading 22C, which is 1C below ambient and 4C below socket ambient: incorrect.

95W TDP CPU under maximum TAT load at lowest fan speed after 15 minutes [Fig.3]:


Now if you understand TDP, you know that 95W on any CPU will give the same end internal core temperature. TAT is extremely high load beats anything in software I've seen. As you can see, Core Temp is reading 54-59C across the cores whilst Real Temp is reading 49-54C. The delta being 5C, which is because Real Temp choose a -5C Tjunction_max. For a 95W TDP CPU or below, sub 50C lowest fan speed air at 24C socket ambient is hard to believe with TAT load. Prime 95 will be at least 3-4C lower across the cores and Orthos even lower.

Here are the temps in all trusted applications at idle, even Intel TAT agrees with Core Temp and the rest of the tools for a 100C Tjunction_max [Fig.4]:


As you can see, Real Temp falls sub-ambient but not the others. Still we have not tested Tjunction_max, so Real Temp may be right although requiring an + offset.

Now you see at 1600MHz 1.152V (not CE1), lowest fan speed, same conditions [Fig.4]:


Real Temp is reading sub-ambient again, none of the rest are.

Drop the volts to 1.120V at 1600MHz, fastest fanspeed [Fig.5]:


Temps have dropped 1C more across the cores, 4C below ambient socket temps.

Time to drop MHz, volts and thus TDP. 1230MHz with 0.992V [Fig.6]:


At those settings and parameters we know, Id constant, the cores are now at 33.3W TDP. Intel states TDP for Q6600 G0 in CE1 is 24W, if that's true, then current is also reduced, which means the new TDP in the above state is 14W, very low. Still, it is now 5C below ambient processor socket temp. on two cores with a lowest speed heatsink+fan.

At 13W TDP, 23C ambient socket temp, took the heatsink off with no fan around. Temperature started shooting up without any load, very quickly, I monitored what the Tcase temp. read with a K-type accurate to +/- 0.3C whilst controlling these to see when exactly PROCHOT# was asserted so that Tjunction_max is reached [Fig.7]:


As you see, everything turned wacko the minute PROCHOT# was reached. But repeated 4 times, the temp. for Tjunction_max to be reached was 100-98-98-100C. At 95C Real Temp reading, the DTS MSR register still read 5C left. Temperatures at this stage are shooting up very quickly without a heatsink. K-type was reading 68C on the IHS now, too hot to touch, burnt my finger by haste moving things around. That's the exact stage I threw a HSF on it to cool it down meanwhile I note things down. You can see the hex MSR values from the DTS in my previous screens, they are inline.

If 100C was Tjunction_max, then it should shut down very soon.. 15-25C after that, right? I tested it [Fig.8]


It was very hard to get an ss before shutdown, temperature runs away at this stage too fast to do anything. 5 attempts got me the above, which is 2C below the temperature it shut down repetitively, without any load, no HSF, 23C socket ambient, THERMTRIP# was asserted. All temperatures in the utilities started reading backwards, only Core 0/1 in Real Temp did not freak out after 110C, and carried on moving up in line. My shutdown temperature in those two readings was 125C. Exactly in line with a 100C Tjunction_max. At the time of shutdown, Tcase on the surface (not embedded) measured 84C with a K-Type.

So I have pretty accurate idle and load temperatures and the MSR was of a big help, as when I reached 95-100C, I checked up what delta it read from the Tjunction_max, to see what was seeming correct. Now I've set 100C TjMax in Real Temp, no offset needed and temps are right on AFAI can tell.

[unclewebb]

I got into a discussion with another user on a different forum and I decided to do a more thorough study of the IR reported temps vs what RealTemp displays. I think the results are very interesting and it also helps to explain the RealTemp Idle Calibration feature which some users don't trust or disagree with.
--------------------------------------------------

Here is my latest guinea pig, an E2160 Revision M0.



Intel has a habit of creating one processor model and then making some slight changes to the basic design so they can create additional models for different price points without having to do a total redesign. The E4x00 series with 2MB of cache gets half of that disabled and then becomes processors like my E2160 with 1MB cache. They then cut the cache in half again and create things like the E1200 M0 which has only 512KB of cache. They all have the same CPUID. Some people believe that this is a way for Intel to sell defective processors. If the cache has a defect, they can disable half or three-quarters of it and still sell it as a lower end model and make some money.

So my theory is that all M0 processors start life as the same basic processor with the same TjMax. As they go down the assembly line, the multiplier and VID and cache are adjusted which creates a variety of models. These different models get stamped and shoved into the appropriate box but beneath the skin they are all quite similar.

GigaByte has brought up some troubling findings that need to be addressed. His M0 processor when using RealTemp's new TjMax=85C is reporting below ambient core temperatures at idle which everyone knows is impossible. When this same situation originally happened with the L2 - E4300 series, the solution CoreTemp came up with was to bump TjMax from 85C to 100C. Changing the TjMax will cure below ambient idle temps but it also screws up your load temps by +15C if the TjMax really isn't 100C. Not good.

The goal of RealTemp was to convert the on chip DTS data to accurate core temperatures. When I first got my E2160 installed and ran RealTemp v2.11, it was immediately obvious that TjMax=95C was wrong for my new processor. I quickly changed that to TjMax=85C for RealTemp v2.2 based on my IR thermometer testing.

With GigaByte strongly objecting to this change in TjMax, I decided I needed to re-do my E2160 testing. I recorded the IR thermometer reading and then immediately after I took a picture of the screen to compare it to what RealTemp was reporting. I'll start with a lot of pictures taken at different temperatures to compare IR gun temps to what RealTemp reports. I have cropped the original pictures down to size to make them easier to upload. I have also kept the original pictures which are all time stamped showing that the video screen shot was generally taken 2 or 3 seconds after the IR picture. It took me a couple of seconds to reposition the camera, refocus and shoot. If anyone is suspicious that my image gathering technique was done to make RealTemp look better then I have the originals that I can e-mail to you so you can check the file creation time stamp on them.

Anyway, here goes. In each group you have the IR measured temp followed by the RealTemp reported temp. Idle Calibration was set to (+) but that feature doesn't effect temperatures in this range.

IR Temp = 50.6C

RealTemp = 50C


IR Temp = 55.4C

RealTemp = 54C / 55C


IR Temp = 65.8C

RealTemp = 67C


IR Temp = 70.6C

RealTemp = 70C / 72C


I think everyone will agree that there is a very strong correlation between IR thermometer measured temperatures and what RealTemp is reporting for core temperatures. The upper range looks good and shows that TjMax=85C for my M0 processor looks like a wise choice.

Now it's time to look at the low end using the same TjMax=85C. With the Idle Calibration set to (+), you'll also see how that effects reported temperatures. I've included pictures of SpeedFan since it uses TjMax=85C for the M0 processors but doesn't have RealTemp's Idle feature so you can compare.

IR Temp = 27.8C

RealTemp = 27C SpeedFan = 25C CPU = 40C


On this processor SpeedFan, without the Idle Calibration feature, is reporting core temperatures that are a little too low. This is common on the original B2 processors, definitely the L2 processors like the E4300 and now also on the M0 processors. The other thing to note is the CPU temperature which is from a diode on the motherboard. On this Asus board, it has nothing to do with the actual TCase temperature of this processor and is completely wrong.

IR Temp = 31.8C

RealTemp = 32C SpeedFan = 29C


IR Temp = 34.0C

RealTemp = 34C SpeedFan = 32C


IR Temp = 41.0C

RealTemp = 40C SpeedFan = 39C


Once again there is a strong correlation between IR temperatures and what RealTemp is reporting. A properly chosen Idle Correction factor is also helping to make these reported temps spot on.

Everything is looking good so far but now it's time to investigate if my idle temperatures are being reported below the ambient temperature like GigaByte has found.

For this test I've reset the Idle Calibration to 0. I turned off my computer in my very cool basement and let it sit. It was left in Stand By mode for a few hours so when I turned it back on I could immediately see the reported core temperature before it had a chance to warm up too much.

I had a notepad sitting beside the computer which I used to record a room temperature of 11.2C.


Here it is. With no Idle correction factor, RealTemp is reporting 8C for the core temperature now.


RealTemp, uncorrected, is reporting this M0 E2160 3 degrees below the ambient temperature. This is exactly the same as what you discovered GigaByte. You were operating your E4500 at 0.90 volts vs 1.08 volts during my testing. Heat output of a processor is proportional to the square of voltage so your CPU was putting out approximately 70% of the heat energy as my processor and one of your cores was reporting that it was at 4C below your ambient temperature.

GigaByte, can you now see how the DTS is not linear. Without an Idle correction factor, these CPUs will report idle temperatures that are too low and impossible. That does not imply that TjMax is wrong. My testing makes it perfectly clear that TjMax=85C is definitely right for my M0 processor and our combined testing shows that our processors are behaving identically by both reporting below ambient idle temperatures.

Your correction to adjust the TjMax upward 10C to 95C will only make all of your load temperatures wrong. For accurate reported temperatures that cover the entire operating range of a processor, you need an Idle Calibration correction factor.

Down at this temperature, RealTemp uses a correction factor of 4 degrees which makes the reported idle temperatures much more accurate.



A 4 degree correction would bring the previous 8C uncorrected reading up to 12C which is above my ambient temperature of 11C. The RealTemp reading may not be 100% correct but it is definitely very close to the actual core temperature of this processor. Closer than any other core tempreature software available. RealTemp, when properly calibrated, is very accurate from idle to TjMax and beyond.
------------------------------------------------------

After all this, GigaByte still disagrees with my program using TjMax=85C for his CPU and bases his arguments on TCase vs core temp gradients that have been proven false by Intel themselves. Oh well, I gave it my best try! Any opinions on this, for or against?

KTE: Thank you for your testing. Looks like I need to buy a G0 Q6600 this week and do some thorough testing myself. Your screen shots are very valuable and the next release of RealTemp will be better because of it.

One thing I noticed about your high temp testing results is that your first screen shot with the Idle Calibration set to (+) shows a big bug in RealTemp when PROCHOT# becomes activated. The Idle Calibration feature is designed to have no effect here but it is screwing up the reported temperatures. I need to fix that.

On your last screen shot you have Idle Calibration set to 0 and now RealTemp is working as it should.

Reported Temp = TjMax - DTS

Some programs like CoreTemp will display the actual DTS values when you go over TjMax and then it puts a ? beside the reading showing that it doesn't know what to do with it. Other programs like TAT will display negative numbers like -40C. This is because the DTS is a 7 bit number and wraps around. My program interprets this as:

DTS - TEMP
2 - 93C
1 - 94C
0 - 95C
127 - 96C
126 - 97C
125 - 98C

etc., etc.

To correct for this, the modified formula after PROCHOT# is reached becomes:

Reported Temp = TjMax - DTS + 128

Your final screen shot shows the DTS at 100 on core 0 and core 1. This means the DTS has wrapped around and you are 28 steps beyond TjMax. My program assumes your TjMax=95C so using the modified formula it calculates that your core temperature is 123C. You state that your CPU then shuts down 2C after that which equals 125C. That is exactly equal to the documented shut down temp for the mobile Core processors.

If your TjMax=100C then your actual core temperature is 128C, here and your CPU is shutting down at 130C. That's beyond the 125C that Intel quotes for their mobile processors. Another sign that more real world testing needs to be done similar to my E2160 test. Controlling an E2160 was simple and boring with no heatsink but the Quad is a different animal. Wish me luck!

[loonym]

KTE, great post! Thank you very much for that explanation and illustration.

[Demo]

Yeah, but you're using the ++ correction factor. The reason 100*C Tjmax was originally selected by most programmers is b/c of idle temp readings relative to ambient. So CT may be indeed be correct at idle, but at load it would be off by ~5*C b/c of an incorrect Tjmax. RT should be correct at load, but needs the correction factor to be accurate at idle. Again it goes back to the theory that DTS doesn't scale linearly.

Idle temp calibration also affects load temps

0 calibration
55-52-56-51

++ calibration
56-55-57-53

EDIT:
To be on the safe side, im opting to use ++ calibration even it affects my load temps.