Real Temp - New temp program for Intel Core processors

Quote:

---

Originally Posted by ixtapalapaquetl

KTE, if any of your talks include technical information, would you mind posting their content here? I am assuming of course that you are communicating through correspondence.

---

Because it's my uncles personal contact (not mine) rather than normal user technical correspondance, I'm limited to not being allowed this liberty, sorry. I know the contact quite well since 2004 and we chat very informally rather than in official tones if I'm around at my uncles work place (I have met and worked with him at times). I can tell you what I learn non-verbatim though, but I ask for personal knowledge and experimentation rather than official responses, bear in mind. I have already mentioned much of what we discussed with respect to temperatures and Core 2 DTS in my last major post with my own Q6600 G0 TjMax hunting, for instance, TDP +3W, IC temp. idle cannot be equal to processor socket ambient temp. on air cooling, it has to be higher, even if slightly.
That's one thing to notice, the CPU MFGs don't look at case and environment ambient temps. alone but the most crucial ambient temp. to them is the socket ambient which has to be lower than the actual CPU temp. but higher than the case ambient temp. in all cases (unless more extreme cooling is used i.e. even a chiller). I also verified the safe rating advised at a personal level, I asked what the maximum voltage he would run without worries at 45nm when overclocking, it's 1.3625V tops safe for 45nm. I was warned to stay below it and not even keep 1.35V volts prolonged on air to be frank (note: Intel is not liable for any damage by my personal comments discussing overclocking!). I also asked what the lowest E8400 temperature would be possible on air cooling, at 22C socket ambient at 1.4V real Vdd supply, stock 3GHz and if 35C idle was possible without CE1 (what I was getting), and the answer was not a chance in definite tones. At 1.4V on air cooling, the core would be very hot, idling plus 40C and the contact doubted highly you can even hold temps. below 45C on air idling at that voltage for E8400. These were for my personal guidance though as I have to troubleshoot and setup many systems at the IT department at two of my workplaces (both scientific, one purely medical), so I take serious advice necessary. :)

unclewebb, I missed the end of your post after my Q6600 testing post. But now that the RealTemp bug is fixed, knowing my TjMax, temperatures above PROCHOT# should show up correctly past 115C, right?

Understood KTE. Muchas gracias for what you have shared. It is nice to hear some corroborating numbers if even in but a semi-official capacity.

Quote:

---

Originally Posted by jas420221

Did you replace the file in the links? I re DL it and its still 1.1. The program when its opened is also 1.1. The notes arent updated either...

---

That seems to be a problem with some web browsers and that site. It finds a copy of the download in the cache and then gives you that one instead of the most recent one on the FileDen server. I just tried downloading using the link in the first post and I didn't have a problem getting the latest version.

In FireFox if I right click on the link and select Save As... and then change the name of the zip file, it seems to download the latest copy without finding something in the cache.

Quote:

---

KTE: I missed the end of your post after my Q6600 testing post. But now that the RealTemp bug is fixed, knowing my TjMax, temperatures above PROCHOT# should show up correctly past 115C, right?

---

I still have the GO Q6600 at a default of TjMax=95C. I'm hoping that I cured all the RealTemp bugs when going beyond PROCHOT# with version 2.21. I guess you get to be the official tester of that. Show us another 120C screen shot so users can stop worrying when they see a core temp of 70C. :eek:

Damn... seems like everyone got it working for Vista 64. All I got is a "driver not loaded" error when I try to run the program. :(

You need to run this program as an Administrator. If not, it won't be able to load the driver. Also make sure the program is fully unzipped.

KTE: I appreciate what you've learned from your Intel contacts but I think sometimes blanket statements are made that aren't based on what a top notch air cooler can do for idle and load temps. Anandtech has done a lot of good air cooler reviews:

http://www.anandtech.com/showdoc.aspx?i=2981&p=3

If your Intel contact is basing cpu temps on using their retail box cooler then there could be a big difference between his numbers and what's possible with an Ultra 120 eXtreme. I also know if you manually drop the core voltage down to 1.10 volts or lower that it also makes a significant difference in idle temps.

does he have the latest version?

I created a copy of the download and renamed it. Try this link if you're having troubles.

http://www.fileden.com/files/2008/3/...lTemp_2.21.zip

Totally forgot the "Run as Administrator" part of Vista. It works. Shows the same exact temps as Core Temp, and Everest without any calibration both idle and load for the E6600 (TJ = 85C) I'm using. The minus/minus and plus/plus calibration will deviate 2-3C at most. So I guess I've got a pretty accurate sensor?

The (++) (--) Idle Calibration feature is mostly for the fanatics amongst us that go ape s-h-i-t when they see an idle temperature that is a couple of degrees below their ambient temperature. Some users see that as a sure sign that TjMax must be wrong but it usually isn't. If you run your E6600 at a very low core voltage of 1.10 volts and underclock it to 1600 MHz (266x6) then it will likely confirm that your idle temps are too low. A (+) or (++) calibration setting is often times best for the early Conroe Revision B2 processors.

I just uploaded a RealTempLogTest file.

It was run on my 65nm E2160. It consists of 4 runs of one minute of idle followed by one minute of Orthos to get some good temp movements up and down so I could compare how the two DTS sensors track each other. I set it to log the temps at an interval of 1 second so there is 478 data points for each core. 98.3% of the time or 470 of the 478 temp readings are identical. The overall average difference between the core readings is only 0.04 degrees. That's a little tighter quality control then the over 30C difference in Brama's cores at idle. Let's hope Intel gets this fixed in the Revision C1 E8x00 dual cores.

Quote:

---

Originally Posted by unclewebb

I still have the GO Q6600 at a default of TjMax=95C. I'm hoping that I cured all the RealTemp bugs when going beyond PROCHOT# with version 2.21. I guess you get to be the official tester of that. Show us another 120C screen shot so users can stop worrying when they see a core temp of 70C. :eek:

---

No problem, give me some minutes to open it up... I'll try PROCHOT# and THERMTRIP# again, but don't want to cut my hands all over the fin 'n' fans when I see 125C again :) :eek:

Quote:

---

Originally Posted by unclewebb

KTE: I appreciate what you've learned from your Intel contacts but I think sometimes blanket statements are made that aren't based on what a top notch air cooler can do for idle and load temps. Anandtech has done a lot of good air cooler reviews:

http://www.anandtech.com/showdoc.aspx?i=2981&p=3

If your Intel contact is basing cpu temps on using their retail box cooler then there could be a big difference between his numbers and what's possible with an Ultra 120 eXtreme. I also know if you manually drop the core voltage down to 1.10 volts or lower that it also makes a significant difference in idle temps.

---

Very true, I agree there could be a massive difference between a decent air cooler and a stock HSF, however, microprocessor and optical lithography engineering isn't a new concept in my family; I'm aware how power used for work within the cores has to be emitted as TDP externally (extremely little is silicon loss) and that sets the IC internal temperature limits even if we could say we know more and he was bluffing me (he's a ~16 year family friend not just a random contact (:D)). You can have the metal surface (Tcase) idle ambient with large air flow and good thermal resistence between contact materials but you can't have the internal core producing the heat idle at or near ambient (i.e. 0W or 0J/s) even at 1.5W unless you had it naked and were directly cooling the core itself with a lot of pressurised air and good contact thermal resistence. Still, it would not produce a core internally idle ambient temps. unless it was off producing 0W TDP, as we measure using the DTS... only the immediate air contact surface could be, as heat emitted cannot be equal to surrounding temps. and carry increased thermal energy. We have to remember basic circuitry and engineering physics: heat (Q), passively will not travel but from a higher temperature subsystem to a lower temperature subsystem. Heat naturally can only transfer between areas and objects with a temperature difference between the two (zeroth law of thermodynamics) and to the colder object/area (second law of thermodynamics).

Quote:

---

Originally Posted by zeroth law

When two systems are put in contact with each other, there will be a net exchange of energy between them unless or until they are in thermal equilibrium, that is, they contain the same amount of thermal energy for a given volume

---

When I tested idle (personally), I did so with a boxed cooler first (I do all my testing with it unless really required), Zalman 9500AT next, to Zalman 9700LED, then Tunic Tower 120 which I used mostly and lastly with what I use with the Q6600 now, a Xigmatek HDT-S1283 with a 152CFM 120mm fan. The idle and load difference between the latter 3 and the TRUE are quite not much, esp. idle, negligible within +/-1C. All are good enough for air cooling over the stock HSF I believe. :)

read through the post and if I have an e8400 I should just leave the calibration in the middle correct? real temp reads exactly 10 degrees less then core temp and everest, due to the 105tjmax and the 95tjmax correct?? Sorry I'm not the best when it comes to this :banana::banana::banana::banana:. unclewebb?? Thanks man..So I should go off real temp numbers not the other programs. :confused:

In testing I'm finding that it's not the cores that are the biggest heat sources at idle but it seems to be the on chip cache memory. I have three different dual core cpus at the moment, one 45nm with 6MB cache, a 65nm with 2MB of cache and a 65nm with 1MB of cache. Without a heatsink on getting the E8400 up to TjMax wasn't a problem while the E2160 was taking forever to warm up and I had to start running a single thread of Prime to get some heat out of that chip. Perfect chip for the home theatre types. You could definitely run it without a fan.

At 1.1 volts and 1200 MHz, my testing seems to show that the E2160 will stabilize at idle at a core temperature about 4C above ambient with a Tuniq on top running at full speed. I wouldn't mind carving up the IHS and putting a thermocouple in the center like Intel recommends but I probably won't.

My Q6600 arrived today but it looks like someone has broken the Intel seal and then the store covered it up with their own seal. I'm not impressed. I have to waste my testing time tomorrow bringing it back to find out what's up.

ban916: I think TjMax=95C is more likely for your E8400. Not too many users here have complained about that number for the E8x00 desktop series. I know my E8400 likes that number. As long as you're stable then your core temp is just a number and you shouldn't worry too much about it.

That's definitely inline with engineering. Cache is supposed to be the major source of heat, taking up most of the transistors. :yepp:

On another note, I tested and not entirely good news. Seems there is still a mishap with reading temps.

Setup
http://picsorban.com/upload/e3-setup.png

Idle, maximum fanspeed (check the system wattage, 47W DC, less than 3W on the 8-pin 2x 12V rails)
http://picsorban.com/upload/e1.thumb.png

Socket ambient is 22-23C, CPU current is negligible (12V1, 12V2):
http://picsorban.com/upload/e2.thumb.png

95C reached, still no PROCHOT# asserted (TjMax not reached):
http://img214.imageshack.us/img214/9...procsn8.th.png

The millisecond 100C was topped, PROCHOT# was asserted:
http://img149.imageshack.us/img149/4...procee9.th.png http://img386.imageshack.us/img386/6...proctg2.th.png

113C, no shutdown yet (no THERMTRIP# assertion):
http://img149.imageshack.us/img149/406/e7113nb8.th.png

Now my hottest core, core0, after 94C is always around 6C hotter than core1. This is what happened when I let it go past 120C, bugged out:
http://picsorban.com/upload/e8-120.png

What you're seeing is, the second either core reaches 120C, it bugs out to 0C readings and 100C from Tjunction_max values. That's 120-121C, a jump up from 118C, and it shuts down if I leave it 2 steps more. Knowing my increases are in 2-4C steps after 94C, the shutdown was most likely at 124C twice repeated today.

It was smelling and making "fizzing" sounds this time round... damn I don't want to lose a 4.26G stock HSF CPU now do I. :(

@KTE,

Your Q6600 can run at 4.26Ghz ??

@Topic

I cant wait for those Q6600 results from you unclewebb so i can have a peace of mine with my correct load temps.

unclewebb
can you add the *percent from tjmax* item to RealTemp?
so when there is no abs temps from dts, maybe this item will be useful.
Like if tjmax=95 and dts shows *50 to tjmax* there will be 45°C or 47% of core's maximum workable temperature.

PS How soon you can add the temps to the tray?

PPS:
My Quad result:
IH-4400A cooler + 2000rpm fan
ambient t=21°C
RT 2.21 shows in idle 30-30-25-29 w/o calibration
so I think there is no calibration needed, right?

Quote:

---

Originally Posted by unclewebb

I created a copy of the download and renamed it. Try this link if you're having troubles.

http://www.fileden.com/files/2008/3/...lTemp_2.21.zip

---

That link did it, Thank you! The ones on the first page are still the old version...

KTE, how are you measuring your absolute temps? When you say "95C reached, still no prochot#"...how do you know you are at 95C? Unless you are measuring temps in an absolute manner independent of the programs, ie thermocouple or IR, you would not.

All your showing is DTS is 5 from zero, so whatever tjmax number you plugged in, you will conclude your absolute temps to be that number minus 5C. Since you set tjmax at 100, at DTS = 5, yes it will show 95C. Had you set tjmax to 95, you would show temps as 90C at DTS = 5, and thus concluded that tjmax is 95...circular reasoning.

Then you state the minute 100C reached, prochot# was asserted, same circular reasoning. All you showed was when DTS=0, prochot# was asserted, and whatever Tjmax you used, will be the temp that is displayed at DTS=0.

You could duplicate the experiment again, using tjmax of 95 in realtemp, and then conclude tjmax is 95 at DTS=0.

Same issue arises at shutdown. Shutdown occurs at 20 to 25C beyond DTS=0. Suppose you knew it occurred at 25C beyond DTS=0. Since you do not know the absolute temp when DTS=0, you do not know the absolute temp when shutdown occurs, you will simply be showing that shutdown occurs 25C after DTS=0 is reached.

EDIT: speaking of intel, I had posted a question up on their forum a month ago when trying to determine gradients, and my post just disappeared, so I posted it again just now.
New link...http://softwarecommunity.intel.com/i.../30251183.aspx
Old link that disappeared...http://softwarecommunity.intel.com/i...eID=6#30250058

Quote:

---

Originally Posted by KTE

That's definitely inline with engineering. Cache is supposed to be the major source of heat, taking up most of the transistors. :yepp:

On another note, I tested and not entirely good news. Seems there is still a mishap with reading temps.

Now my hottest core, core0, after 94C is always around 6C hotter than core1. This is what happened when I let it go past 120C, bugged out:
http://picsorban.com/upload/e8-120.png

What you're seeing is, the second either core reaches 120C, it bugs out to 0C readings and 100C from Tjunction_max values. That's 120-121C, a jump up from 118C, and it shuts down if I leave it 2 steps more. Knowing my increases are in 2-4C steps after 94C, the shutdown was most likely at 124C twice repeated today.

It was smelling and making "fizzing" sounds this time round... damn I don't want to lose a 4.26G stock HSF CPU now do I. :(

---

very nice, thorough tests and write up.:up:

When I decided to create a temp program for the XS crowd I thought if they could get accurate temps up to damn near 120C that they'd finally be happy but of course not. They want to go higher! :yepp:

KTE proved that version 2.21 was crapping out near 120C so for anyone that needs to go higher then download the latest version 2.24.

http://www.fileden.com/files/2008/3/...7/RealTemp.zip

http://www.fileden.com/files/2008/3/...lTemp_2.24.zip

It should be good for 145C now. ;) Cash reward for the first person to send me a screen shot of that without using PhotoShop. Reward does not include a new processor. If this latest fix causes any problems with people exploring the sub zero spectrum then send me a screen shot so I can look into it.

I believe KTE that when the DTS was reporting 100 when you were well beyond TjMax that if your TjMax is 100C and if the DTS data up here is still reasonably accurate then your processor was at 128C and if your TjMax=95C like I believe then you were at 123C. I think things are a little too crazy up here as well as undocumented to really prove anything one way or the other though. The only thing really proven is that the 65nm chips can take a hell of a lot of abuse without crapping out. I don't think I'd be trying the same with the new 45nm chips quite yet, especially a QX9650.

My testing and development time is going to be nil for the next week. Testing a Q6600-GO and the Minimize to Tray option are the most important things on the Things to Do List.

rge: Looks like your previous question went into Intel's mysterious black hole of knowledge. I won't hold my breath waiting for them to provide you with an answer. I can't imagine that they want to talk too much right now about anything to do with the data coming out of the DTS sensors, especially in the Revision C0 45nm chips.

Unclewebb - May I suggest (again) that when you update the program you update the links on the first page...

As you can see this is a HUGELY popular thread with almost 500 posts and its very tough to find the latest version if its not cleary updated somewhere in the first post. Thanks again for your efforts!!!!

PS- Any idea on when you can get tray temps to show up? That would be HUGE!

jas420221: As I said before, the link to the latest version in post#1 never changes. If you click on it and get an old version it is because your computer has an old version stuck in its cache. Clear the old version out of your cache if you are ever having problems. Doing it this way makes it easier for other sites that have linked to it so the links don't have to be updated every time I come out with a new version and we all know how often that has been lately!

If you want tray temps RIGHT NOW then do what I do and use SpeedFan. The RealTemp code for this is still a work in progress. I think I spent too much time working on Brama's custom version for his messed up E8400 this week so blame him. :D

Quote:

---

Originally Posted by rge

KTE, how are you measuring your absolute temps? When you say "95C reached, still no prochot#"...how do you know you are at 95C? Unless you are measuring temps in an absolute manner independent of the programs, ie thermocouple or IR, you would not.

All your showing is DTS is 5 from zero, so whatever tjmax number you plugged in, you will conclude your absolute temps to be that number minus 5C. Since you set tjmax at 100, at DTS = 5, yes it will show 95C. Had you set tjmax to 95, you would show temps as 90C at DTS = 5, and thus concluded that tjmax is 95...circular reasoning.

You could duplicate the experiment again, using tjmax of 95 in realtemp, and then conclude tjmax is 95 at DTS=0.

Same issue arises at shutdown. Shutdown occurs at 20 to 25C beyond DTS=0. Suppose you knew it occurred at 25C beyond DTS=0. Since you do not know the absolute temp when DTS=0, you do not know the absolute temp when shutdown occurs, you will simply be showing that shutdown occurs 25C after DTS=0 is reached.

---

You're making some good points there, even though they are things I'm well aware of. I'll explain fully but not before much more experimentation needed and I need more time now since my system is in a case used every day and I don't plan on dismantling it again now having just put it back, I have much else to do. It is obviously flawed to "guess" TjMax, since as you mention, Real Temp will only show TjMax you set minus DTS as your core temperature (without offsets).

Everyone else should pay due importance to what rge mentions here as it applies to each of us. If I set TjMax to be;

70C, my DTS reading 5C will give me 65C core temp. in Real Temp.
75C, my DTS reading 5C will give me 70C core temp. in Real Temp.
...
90C, my DTS reading 5C will give me 85C core temp. in Real Temp.
95C, my DTS reading 5C will give me 90C core temp. in Real Temp.
100C, my DTS reading 5C will give me 95C core temp. in Real Temp.

The issue arises when in cases like mine;
*If 100C TjMax gives you absolutely no error in temperature readings at idle, then give me one plausible reason to change it to a lower or higher value?
*Now, if changing to 95C TjMax only makes your temperatures 3-4C less than ambient at ~8W TDP (impossible) compared to 1-2C above ambient for 100C TjMax, why would anyone change that unless they're after getting wrong temperature values? (i.e. lower) ;)

*Actual TjMax value is fixed, you can't change it for a CPU.
*Actual PROCHOT# assertion value is fixed, you can't change it for a CPU.
*The same applies for THERMTRIP#, the difference is only caused by feedback delay causing Vcc and Vss cut off to shut the CPU down.

The case becomes even more convoluted when;
*You measure the Tcase temperature has nothing to do with DTS Tjunction temperature reports (similar to what Intel measured across different work loads and different software, Tdiode vs Tjunction (DTS) -> Fig.5), coincidentally, that is very similar to what I measured with all my last E6750/E6850/Phenom 95/96/96BE/Q6600 B3/G0 testings when I kept them and some I still have. I found the Tcase did not equal Tjunction and the gradient between the two in some loads and temperatures was much less than at other loads and temperatures. Usually at low Tjunction temperatures idling at low TDP, the delta was +/-1-2C, but at higher TDP, temperatures or loads, anywhere up to +/-28C I've measured**.
*Tcase_max Intel lists as 71C maximum permitted for my Q6600 G0 at 95W maximum TDP.. my CPU delta from Tcase on the geometric surface center (not embedded) with a K/N-Type thermocouple, from internal Tjunction DTS reports anywhere from -1C to -28C throughout a range of TDPs. Based on this data collection, how does Tcase become relevant to Tjunction for my CPUs which show this, or how can we correlate the two?

**Measuring Tcase at the geometric surface center like I did is still not 100% accurate. The way Intel has instructed is a very careful procedure to embedd the thermocouple within the case (IHS) to attain accurate Tcase temperature readouts or the temperatures can be very inaccurate: [Page 32 and Page 85, Appendix D Case Temperature Reference Metrology]

Quote:

---

Originally Posted by Intel

To ensure functionality and reliability, the processor is specified for proper operation when TC is maintained at or below the thermal profile as listed in the datasheet. The measurement location for TC is the geometric center of the IHS. Figure 2 shows the location for TC measurement.
Special care is required when measuring TC to ensure an accurate temperature measurement. Thermocouples are often used to measure TC. Before any temperature measurements are made, the thermocouples must be calibrated, and the complete measurement system must be routinely checked against known standards. When measuring the temperature of a surface that is at a different temperature from the surrounding local ambient air, errors could be introduced in the measurements. The measurement errors could be caused by poor thermal contact between the junction of the thermocouple and the surface of the integrated heat spreader, heat loss by radiation, convection, by conduction through thermocouple leads, or by contact between the thermocouple cement and the heatsink base.

---

Quote:

---

Originally Posted by rge at Intel forum

EDIT: speaking of intel, I had posted a question up on their forum a month ago when trying to determine gradients, and my post just disappeared, so I posted it again just now.
New link...http://softwarecommunity.intel.com/i.../30251183.aspx
Old link that disappeared...http://softwarecommunity.intel.com/i...eID=6#30250058

---

Good qs, It'll be interesting to hear what they say too, at a public official level. They've had many issues with developers before on this topic and this has all been discussed by developers before with Intel, they hardly mention anything. Everything they mention officially is all here really: http://softwarecommunity.intel.com/i...howThread.aspx

Quote:

---

Originally Posted by rge at Intel forum

I had asked this question before, but can not find my post. Just trying to ascertain temps of my cpu.

For an E8400, at idle, equilibrated steady state with no heatsink, underclocked 1200mhz (6x200mhz) and 0.9v vcore, measuring tcase as 75 when DTS=20, and 95 when DTS=0, can one conclude that the gradient from tjuncion to tcase is ~0-2C, and thus Tjmax for that cpu is 95C +/-2C?

---

So we're talking near to 4W TDP I suppose, if not less. Tcase and Tjunction should be very close at low temperatures but I'm not sure as heat builds up how that would show. Specifically I am not sure if Tcase can equal Tjunction in any such scenario, when Intel lists the maximum permitted safe Tcase for all cooling manufacturers to develop cooling on as 72.4C at 65W but you list as having measured ~95C at ~4W: http://processorfinder.intel.com/det...x?sSpec=SLAPL#

As you can see in Intel documents [Page 77, Figure 13], maximum Tcase temperature is a function of the maximum CPU TDP; it's a linear correlation.

Code:

---

TDP = 0.42.Tcase_max + 45.1

---

Therefore, the lowest maximum Tcase temperature is 45.1C at 0W TDP.

Quote:

---

Originally Posted by Intel

This table shows the maximum TDP for a given frequency range. Individual processors may have a lower TDP. Therefore, the maximum TC will vary depending on the TDP of the individual processor.

---

-At 65W the E8400 has a maximum Tcase permitted temperature before damage at 72.4C.
-In CE1 Intel states the E8190/E8200/E8400/E8500 as 8W TDP at 36C Tcase temperature and minimum voltage loadline [Page 76, Table 25].
-Now, the ~4W TDP (I haven't calculated because I don't know the Idd your CPU pulls at idle, but it is very near 4W) you have on your 0.9V.1200MHz E8400, means your new Tcase_max is much lower.
-At near 4W TDP, your underclocked/undervolted E8400 has a maximum Tcase_max of 46.8C at 4W TDP and very likely even lower.

Looking at the obvious, does that mean the 95C Tcase you've been measuring is actually damaging your CPU massively, as in 40C above rated maximum Tcase tolerance?

You should also list to Intel in the above post you made how you acquired your Tcase temps. because your/my method is not Intel or scientifically accepted as giving the accurate Tcase temp., nor the method which cooling manufacturers have to use to develop cooling for a certain CPU at a specific TDP. The parameter they use to test cooling is not Tjunction on desktop/server CPUs, as it's not disclosed but the above Tcase_max for each CPU specified by Intel, which is disclosed for IHS containing CPUs.

Quote:

---

Originally Posted by Intel

Thermal Specification: The thermal specification shown is the maximum case temperature at the maximum Thermal Design Power (TDP) value for that processor. It is measured at the geometric center on the topside of the processor integrated heat spreader.

---

Quote:

---

Originally Posted by rge at Intel forum

I know the gradient from tjunction to between the cores, tdiode, is near 0C under those conditions...see article here from intel, esp figure 5
http://arxiv.org/ftp/arxiv/papers/0709/0709.1861.pdf

---

Tjunction (DTS) to Tdiode at various application loads is shown as very high to none with different applications there.

Quote:

---

Originally Posted by Figure 5

It can be seen that large temperature gradients exist on the die. It also can be noted that some workloads display high temperature gradients while other have no offset. Thermal control algorithms need to prevent the hot spot from exceeding the max temperature specification.

---

Quote:

---

Originally Posted by rge at Intel forum

The gradient across the copper IHS is going to be equal to or less than across the die, as it has a thermal conductivity of at least that of the die, thus that gradient would be significantly less than 1C as well under those conditions.

---

The gradient across the IHS does exist (measure edges from center) but less so than the die because the purpose of the nickel plated copper IHS is used to spread the hotspot temperature across the IHS evenly, so it makes good contact with the HSF base, as even as possible is what they want for most dissipation.

Quote:

---

Originally Posted by rge at Intel forum

The remaining question...What is the gradient across TIM1/interface. One intel paper suggests a mold compound ~150 micons thick with a thermal conductance of 0.6W/M*K has a gradient of 0.1C, ie from 68C to 67.9C when 1.2W(0.3x4) applied. More modern die attach adhesives have thermal conductances of 10x that, so even for TDP significantly greater including an idle E8400 under conditions listed above, only about ~1C or less gradient should be incurred. But need to test it, which i am sure intel has.
http://www.flomerics.com/flotherm/te...apers/t324.pdf

---

I don't understand what makes you assert this; have you tested or calculated it, or is this a guesstimate?

Not agreeing or disagreeing with you necessarily, but I would like to know how you know this. Also, your CPU TDP in the above conditions is not far off from what they tested, although at idle -- it may even be 1W idling without load unlike what they had.

Quote:

---

Originally Posted by rge at Intel forum

3) And additionally, one intel doc (a) below states E8xxx has diode and another (b) states E8xxx only has DTS and no diode? So does E8400 have diode, and does capable chip of m/b then read that diode as cpu temp?

(a) "The processor has only DTS and no thermal diode. The TCONTROL in the MSR is relevant only to the DTS" quote from page 39, under graph of http://download.intel.com/design/pro...x/31873401.pdf

versus conflicting intel doc
(b) "These parts have Tdiode enabled" (E8400)
http://processorfinder.intel.com/det...px?sSpec=SLAPL

---

Same 31873401 document shows Core 2 Duo has no diode but LGA775 CPU's before that did:

Quote:

---

Originally Posted by Page 69

Figure 28 shows the major connections for a typical implementation that can support processors with Digital thermal sensor or a thermal diode. In this configuration a SST Thermal Sensor has been added to read the on-die thermal diode that is in all of the processors in the 775-land LGA packages shipped before the Intel® Core™2 Duo processors.

---

Quote:

---

Originally Posted by unclewebb

When I decided to create a temp program for the XS crowd I thought if they could get accurate temps up to damn near 120C that they'd finally be happy but of course not. They want to go higher! :yepp:

KTE proved that version 2.21 was crapping out near 120C so for anyone that needs to go higher then download the latest version 2.24.

http://www.fileden.com/files/2008/3/...7/RealTemp.zip

http://www.fileden.com/files/2008/3/...lTemp_2.24.zip

---

Well, you did ask me to test it saying CoreTemp/EVEREST bug out above Tjunction_max whilst Real Temp doesn't. So I tested it. :p:

Testing safety shutdown temperature for me is a must, they have to work. I do it on every CPU at work when they purchase it (or the others do it). Sadly, I doubt I'll test it again unless I get some answers from Intel on the delta between Tjunction and Tcase at sub 8W TDP idling, because unless we can verify this, all the testing is going down the drain. :(

Quote:

---

It should be good for 145C now. ;) Cash reward for the first person to send me a screen shot of that without using PhotoShop. Reward does not include a new processor. If this latest fix causes any problems with people exploring the sub zero spectrum then send me a screen shot so I can look into it.

---

So it definitely shows subzero? I can test that soon though. ;)

Quote:

---

I believe KTE that when the DTS was reporting 100 when you were well beyond TjMax that if your TjMax is 100C and if the DTS data up here is still reasonably accurate then your processor was at 128C and if your TjMax=95C like I believe then you were at 123C. I think things are a little too crazy up here as well as undocumented to really prove anything one way or the other though.

---

I would say that too if I hadn't been there watching it go up with my own eyes; 102, 106, 110, 113, 116... 0. :D You're assuming 128C because DTS is reporting 100C (28+TjMax), while I was watching the DTS/temps go up with my own eyes to know what it is to be plausible.
The second 120C is reached after 116C, the software+DTS bug out at 0C readings. 116C won't jump to 128C in a millisecond, look at the delta with other cores and compare it to when 5-15C below, it is maximum 6C and minumum 4C. :yepp:

Still, thanks for the software update... Intel TAT awaits me. :up:

I just tired using this program. Can someone tell me why my temps are so different from everest 4.50.1330? isn't the tjmax suppose to be 105C for the qx9650?

Quote:

---

Originally Posted by unclewebb

jas420221: As I said before, the link to the latest version in post#1 never changes. If you click on it and get an old version it is because your computer has an old version stuck in its cache. Clear the old version out of your cache if you are ever having problems. Doing it this way makes it easier for other sites that have linked to it so the links don't have to be updated every time I come out with a new version and we all know how often that has been lately!

If you want tray temps RIGHT NOW then do what I do and use SpeedFan. The RealTemp code for this is still a work in progress. I think I spent too much time working on Brama's custom version for his messed up E8400 this week so blame him. :D

---

Sorry I missed that... my fault. BIG thread!!!

Take your time on that tray temp coding... DAMN YOU BRAMA!!!! :ROTF: