Real Temp - New temp program for Intel Core processors

[unclewebb]

Speedfan 64/64
That program -21/-21 (Tjuntion 85)

Speedfan shows correct temp.

That little program shows the actual DTS and puts a negative sign in front of it for some reason. TjMax = 85C so 85 - 21 = 64C which is what SpeedFan reports.

The code is included so I'm hoping I can switch to this more up to date driver and build a second version with x64 support.

[Ace-a-Rue]

Thanks for the info.

Can anyone confirm if this program works in Vista or XP x64,
http://shefot.com/toni/fajlovi/c2dte...inst_e_x64.msi

It uses the CrystalCPUID driver and might be the easiest way to share RealTemp with the rest of the world.

it works in vista64 but the temps are in minus degrees and way far away from what is actual.

[Ace-a-Rue]

Originally Posted by leffe
Confirmed. Although it shows idle temps around 60 when I'm cooling with water and U-guru shows ~30. Using Vista64.
--------------------------------------------------
Originally Posted by unclewebb
Good news. That program is displaying the actual DTS reading and is accurate.
--------------------------------------------------

hey, unclewebb...did i read the above para correctly?!...Uguru chip on Abit mobo's read from the DTS sensor?

unclewebb...would you be so kind to answer my question about Uguru and DTS sensor?

[fornowagain]

Nice work, just a question.

So a Tjmax is assumed as a fixed generic value for all chips on that particular sSpec? And the absolute values from that.

Isn't the set point (DTS=0) for PROCHOT# TCC individually calibrated per chip? Referenced to the Tjmax at workload hotspots. Meaning that each chips TCC set point could be potentially slightly different? Obviously within a range.

Just wondering about validity of absolute values. Would it be impractical to allow recalibration of the assume Tjmax within you app for advanced users? Or is the generic value so close in practise to the factory TCC to make it unnecessary. Maybe a larger sample is needed to answer that?

[jas420221]

WOrks!

So is this accurate or not? Is there any way to truely confirm its accuracy? As an overclocker, I would want the warmer temp posted so I dont kill my prcoessor. According to this thing I have a long way to go with my processor as far as temps go.

[jason4207]

Of course the DTS sensors are very close to each other. The core is about as big as your thumb nail so the large gradients that used to exist in the Pentium 4 era have been reduced to next to nothing. If one core at full bore brings the other core up to within 1C of it then at idle the difference between an IR gun IHS reading and the "real" core temperature can't be greater than that.

You don't have to go all the way up to TjMax when testing. At an IR reading of 60C, the DTS shows 35 on my E8400. at 70C is was showing 25 and at 80C it was showing 15. I plan to do some more high temperature testing when I get a chance to make sure the PROCHOT# feature is working properly.

OK, but these 45nm chips are made w/ some exotic materials that may transfer heat better. Plus, the cores are physically one very small compact item, and the heat may transfer well between them (or at least the sensors), but that doesn't mean the heat will transfer just as well through the TIM and IHS (which has a much larger surface area, and a much better ability to cool more quickly).

Using your above data, and assuming a 10*C gradient, then the 60*C reading would correspond to a 70*C die temp, and w/ a 35 DTS reading that gives you 105*C Tjmax. It works the same at the higher readings.

Now the lower readings have to be adjusted as you say, but these upper readings don't prove to me that Tjmax is 95*C...quite the opposite.

[unclewebb]

Originally Posted by leffe
Confirmed. Although it shows idle temps around 60 when I'm cooling with water and U-guru shows ~30. Using Vista64.
--------------------------------------------------
Originally Posted by unclewebb
Good news. That program is displaying the actual DTS reading and is accurate.
--------------------------------------------------


unclewebb...would you be so kind to answer my question about Uguru and DTS sensor?

What I was saying was that little utility program is reading the DTS data correctly. It is displaying it as a negative number but it's working. I'm not sure about your Uguru program. I thought those were reading inaccurate motherboard diodes.

Some good news to report tonight. Can anyone see what's missing in this picture?



Say Good Bye to the IA32 driver and the rather complicated install procedure. I integrated that open source driver that was recommended by FELIX and it's working great in x32. I need to re-write some of the processor recognition code which relied on the IA32 driver but RealTemp has been converted over and is reading the temps without the IA32 driver being needed.

Can anyone say, x64 beta testing for tomorrow.

jason4207: It doesn't really matter what the true core temperature is. If you believe that there is still a 10C gradient that is OK. What my program accurately tells me is that with my E6400 when the DTS returns zero, if I point my IR thermometer at the core I get a reading of exactly 85C. When I do the same experiment with my E8400 it says 95C. Real Temp is the only available software that lets me compare temperatures across these two processors. It's also correctly setting TjMax=85C for the L2 processors. SpeedFan has just jumped on the TjMax=100C bandwagon along with CoreTemp for these processors but it's wrong. With TjMax=100C everyone has been thinking that their L2 are running hot. Same with my E8400. CoreTemp using TjMax=105C for an E8400 and TjMax=85C for my previous E6400 results in an error of 10C that can be measured. That's all that's really important to me.

On a side not, how do people's E8x00 processors compare heat wise to their previous Core2Duo E6600 or E6700. Does everyone still believe that 45nm is hotter?

[jaredpace]

I used to push my e8400 to 88C at load. It would get this after only about 20 mins in Prime on coretemp. Does that mean it was actually 78C if realtemp is correct?

Also slightly OT, whats more dangerous: high temperatures, or high voltage? Is priming at 80C-90C okay?

[HDCHOPPER]

pent D not supported

[Retrolock]

On a side not, how do people's E8x00 processors compare heat wise to their previous Core2Duo E6600 or E6700. Does everyone still believe that 45nm is hotter?

Thanks for the program unclewebb! This 45nm chip is definitely cooler. My E4300 was like 41-44c always on idle and was awful at load, atleast 55-60c on 2400mhz. Now, with the e8400 overclocked to 3.6ghzm I'm idling at 33-35c with 45-49c on load.

[scwam]

This thread is to complex for my little brain. Here are my observations since I've got a micro-ntc thermistor (accuracy +/- 1c) under my 8400, touching the bottom/under core against the resistors connected to a fan controller with temp readout.

After 17 minutes orthos (cpu on water):

Thermistor reads 49.1c
this program 51 - 53c core 1 / 47 - 49c core 2

4140mhz at 1.380 post vdroop.

nothing done with program/no calibartion are settings changed.

[i43]

sorry i don't know is this info useful to You ?
here some screens of my system, anyway this seems a lil' complex to me either, but OK no problem, ready to help and share my inf and readings

MBO: Maximus Extrem / Bios v. 1001 / Air cooling / (No water on NB)
CPU: QX9650 [ Fpo: 742A758 ] (vCore 1.2375v manual in Bios) / Air cooling / TRUE120 / Skytce Minebea 120mm /1600rpm
RAM: CellShock (MSC CS3222580) Air cooling / OCZ XTC Ram cooler)
/ case : Antec P182 / side panel open
opsys : Win XP Pro Sp2


both screens was run with Realtemp 1.1 (with ia32 driver installed )
fully aircooled system ! ambient Temp 19~20 C* (measured by Mobo ext. probe ) seen on Everest(window) screen as : Temperature #2 ;
and HW monitor (window) screen as : TEMPIN2
C1E ; CPU TM Func. enabled /

... Idle ...................... Load

[rge]

Nice work, just a question.

So a Tjmax is assumed as a fixed generic value for all chips on that particular sSpec? And the absolute values from that.

Isn't the set point (DTS=0) for PROCHOT# TCC individually calibrated per chip? Referenced to the Tjmax at workload hotspots. Meaning that each chips TCC set point could be potentially slightly different? Obviously within a range.

Just wondering about validity of absolute values. Would it be impractical to allow recalibration of the assume Tjmax within you app for advanced users? Or is the generic value so close in practise to the factory TCC to make it unnecessary. Maybe a larger sample is needed to answer that?

Yes, intel states in multiple places that desktop cpu's are individually calibrated, thus the absolute temps should be slightly different chip to chip.

Interestingly, though with two chips reading 95C Tcase when DTS=0, and based on previous chips readings, makes it likely that this chip temp difference is not much, though perhaps we should call it "average Tcc activation temp", instead of "tjmax", for desktop cpus.

Regarding individual calibration....I would not even if it were possible. Assume for arguments sake that tjmax is set at the best average tcc. Then when chip A (assume -1C difference tjmax) and chip B (assume +1C diff. from Tjmax) are both reading 70C, then both chips are exactly the same temp difference from some identified intel issue. In other words one chip can tolerate 2C more temp, but, intel has fixed it so when both chips read the same absolute temp, those temp readings mean same thing chip to chip. Why lose the beauty of that? Using DTS calibration is good, we just want to use it in the more familiar way, expressed in absolute temps, even though those "absolute temps" may in actuality be a couple degrees different chip to chip.

[jas420221]

On a side not, how do people's E8x00 processors compare heat wise to their previous Core2Duo E6600 or E6700. Does everyone still believe that 45nm is hotter?

My good friend has a 780i, an e8400 and Tuniq 120. His volts are 1.37 (cpuz load) @ 4Ghz. Coretemp showed max load at 63C. This shows 54C. His e6750 on the same mobo and cooler with core temp was 54C load with a .2 volt increase to 3.8GHz.

[Brian MP5T]

Awesome!

[my87csx481]

This is the first public release of Real Temp which is a temperature monitoring program designed for all Intel single Core, dual Core and quad Core processors....

Is there a list of supported CPUs somewhere? I'm on a P4 HT at work....would be nice to know temps. Damn dell....Coretemp doesn't work at all (wrong cpu architecture) and speedfan doesn't find any sensors but on the hard drive.

I just thought I'd ask before I tried to install it. Thanks.

[KTE]

unclewebb: Thanks for the utility, it is much more near to accuracy now. These fudged readings were becoming so common and a false basis of pointless arguments. My E6750 and E6850 would read sub-ambient at 1.4V 3.7G ac too. I tested exactly similar to what you did a while ago with software and a n-type probe and the differences are very similar on both sides, Intel C2/Penryn and AMD Phenom. It's quite odd, I tested in near exact same methods. The hands on method is appealing to me.

However, one thing: the TJmax is not consistent for all CPUs of the same SKU. My present Q6600 G0 throttles at 92C TCASE but my older Q6600 G0 throttled at 83C TCASE under the same conditions (no heatsink/fan). Bear that in mind.

Secondly, I've not spent much time reading 45nm docs yet, but is TJMax internal core or TCASE temp? It's supposed to be TCASE IIRC then it'll obviously be much lower than internal core temp under load, quite easily 15-20C lower. At idle (depending on volts/Speedstep) it won't be far off.

Thirdly, which MSR is it for the DTS?

Is 45nm hotter? Only the higher end quads IME, the TDP shows which is hot. The increased cache and current is what will be making them hotter. The 9770/9775 are Pres-hot all over again without a doubt. You don't need no program to know this, just load them up at stock with stock HSF and watch them throttle inside seconds.

Sorry, very busy to actually delve into the coding right now and check it out. I'll test it later on as I already know my CPU TJmax i.e. exactly at what TCASE it begins to throttle.

[Kenski]

Just finished running the promised half power test,
Here's the results:
QX9650, watercooled (Storm G5, Swiftech 665 pump and mcp320 rad)
Asus P5E (bios 602)
WinXp
Prime95 v25.5 running small fft test
Room temp/rad intake temp =27c
333 x 12 readings @ 1.336v vcore under load:
Coretemp 66-64-64-64
Realtemp 56-54-54-54
333 x 6 readings @ 1.336v vcore under load: (had to trim the vcore a little
as there was less full load droop at the lower clock speed)
Coretemp 52-48-50-50
Realtemp 42-38-40-40
So, starting with Realtemp I took a rough "average" of 54c for 333x12, and 40c for 6x333, with everything being equal apart from the core speed there should be an ideal of 50% power reduction, so,
x6 temp minus room temp should be half of x12 reading minus room temp:
54 - 27 = 27
40 - 27 = 13
pretty close to the theoretical 50%, now Coretemp, I used an "average" of 64 and 50, so:
64 - 27 = 37
50 - 27 = 23
way off the mark for the QX Cpu
Looks Like Realtemp is a keeper for this rig.
Iidle temps don't seem to be accurate though, with speedstep and other powersaving options turned off there was a temp difference of 1c at most between 6x and 12x core speed at idle.

[unclewebb]

Iidle temps don't seem to be accurate though, with speedstep and other powersaving options turned off there was a temp difference of 1c at most between 6x and 12x core speed at idle.

Thanks for the interesting numbers.

I've also noticed that power output (heat) at idle is barely effected by MHz. It's core voltage that creates the biggest change in heat. Power is proportional to the square of voltage so that's where you'll see idle temps jump. If you use the appropriate Idle Calibration factor, your idle temps should be very accurate and more accurate than any other program is displaying.

We need more people to try the 1600 MHz, 1.10 core voltage test as explained in the first post and to post their results. That's the easiest way to improve your idle temp accuracy.

However, one thing: the TJmax is not consistent for all CPUs of the same SKU. My present Q6600 G0 throttles at 92C TCASE but my older Q6600 G0 throttled at 83C TCASE under the same conditions (no heatsink/fan). Bear that in mind.

What software were you using to determine that your processor was throttling? The other question is are you absolutely sure that both of your Q6600 processors were G0? My program assumes TjMax=85C for the early Q6600 - B3 and TjMax=95C for the Q6600 - G0. That explanation would fit your results perfectly if your early processor was actually a B3.

In terms of throttling, I've also done some testing and noticed throttling beginning 2C or 3C before TjMax or DTS=0 but I don't have 100% confidence in the software I was using at the time. This goes against Intel's documentation for their desktop processors. I definitely plan to do some more testing in the near future and post some screens of a hot throttling processor. One other thing, I designed Real Temp to accurately display temperatures beyond TjMax. Neither SpeedFan or CoreTemp can do that.

[Darth Vader]

unclewebb
How to make calibration for four cores?

[unclewebb]

unclewebbHow to make calibration for four cores?

I'll add that to the things to do list. Are your core temps similar at full load? You should run Prime with small FFTs to balance the heat output. Two instances of Orthos, small FFTs would also work. I know that program brings the heat up equally on the dual cores.

At idle on quad cores I usually see two cores running at one temperature and the other two cores running at a different temperature but there doesn't seem to be a pattern of what two cores are running at the same temp.

KTE: Hidden somewhere in post#1 is that info. MSR 0x19C contains the temp data for any Intel Core based processors. The DTS data is 7 bits long and is in bits[22..16]. Other info in that MSR are flags for thermal throttling which I'm hoping Real Temp is going to show correctly.

[Kenski]

Thanks for the interesting numbers.

I've also noticed that power output (heat) at idle is barely effected by MHz. It's core voltage that creates the biggest change in heat. Power is proportional to the square of voltage so that's where you'll see idle temps jump. If you use the appropriate Idle Calibration factor, your idle temps should be very accurate and more accurate than any other program is displaying.

We need more people to try the 1600 MHz, 1.10 core voltage test as explained in the first post and to post their results. That's the easiest way to improve your idle temp accuracy.



What software were you using to determine that your processor was throttling? The other question is are you absolutely sure that both of your Q6600 processors were G0? My program assumes TjMax=85C for the early Q6600 - B3 and TjMax=95C for the Q6600 - G0. That explanation would fit your results perfectly if your early processor was actually a B3.

In terms of throttling, I've also done some testing and noticed throttling beginning 2C or 3C before TjMax or DTS=0 but I don't have 100% confidence in the software I was using at the time. This goes against Intel's documentation for their desktop processors. I definitely plan to do some more testing in the near future and post some screens of a hot throttling processor. One other thing, I designed Real Temp to accurately display temperatures beyond TjMax. Neither SpeedFan or CoreTemp can do that.

True, idle should be a very low power state, plus the good watercooling,
looking at the idle temp reading now it is actually pretty close, 27 - 28c with a room temp now of 25c, just an odd reading from core 1 of 21c is more of an annoyance than a wrong reading, reads the lowest under load too, but only a degee or two.

[Darth Vader]

Are your core temps similar at full load?

the last core is the coolest, like that 50-49-49-42.

[Ace-a-Rue]

Unclewebb:

can i assume coretemp is reading approximately 10C higher than your program, realtemp?...reason i ask, i can get coretemp to work in vista64...just wanted to know so i can deduct 10C from those output readings to know where the actual temps are at.

i saw that you incorporated the IA32 driver into your program...my question: did you update your downloadable zip file or is it still the old way?

[KTE]

What software were you using to determine that your processor was throttling?

Usually set many open to test on any config, then use whichever I find accurate. The ones working were RMClock and Throttlewatch but I'd then start up ChckCPU32 to cross-check the values.

I had to crosscheck the relative values using a benchmark by just running a SPi 512K bench to compare.

The other question is are you absolutely sure that both of your Q6600 processors were G0? My program assumes TjMax=85C for the early Q6600 - B3 and TjMax=95C for the Q6600 - G0. That explanation would fit your results perfectly if your early processor was actually a B3.

Yep, it was early batch G0 step and was running good 12C hotter than the G0 I have now on the exact same setup and cooling (VID difference is apparent). I've never had that one reach past 85C IHS without instant shutdown (when I primed it without HSF). That would be higher internal though.
Cross checking CPUs against one another or against two different ones is only an engineering possibility because you have to determine the current the cores are pulling, the end TDP, the loadline voltages, fixed TJmax, relative TCase, ambient processor temp and work out the thermal resistance of the IHS. You'll also have to know which parts of the core are off and which on, which differs. Penryn is optimized in CE1 state to switch off much of it's logic, hence the power saving/low idle TDP. It's a complex and long winded task with too much room for error.

Like the two Phenoms me and another user here have. Absolutely near exact systems but for two small components (similar power rating) and his is pulling 30W AC idle and 70W AC load more than mine at the same config at 800MHz 0.850V idle. We checked MSR Idd values and his CPU current set at boot is more than mine even at the same MHz and volts. Obviously his TDP will be more and his power draw now, hence higher real temps.

In terms of throttling, I've also done some testing and noticed throttling beginning 2C or 3C before TjMax or DTS=0 but I don't have 100% confidence in the software I was using at the time. This goes against Intel's documentation for their desktop processors. I definitely plan to do some more testing in the near future and post some screens of a hot throttling processor.

I'll test my Q6600 again and an E8400, just waiting till they send me a new batch. Availability here is still very poor and they're getting too many RMA's. I returned mine as it read 14C sub-ambient idling and one core was stuck to -2C always. But even with the lack of time and me working on other systems, I should be able to test throttling temps for the G0 I have now by tomorrow.

4 instances of Intel TAT at 100% for a 450FB quad at 1.40V on air 20C ambient is enough to reach plus 90C CoreTemp temps anyway. I'll throw lowest volts/speeds I can manage idling without HSF, if that reads subambient when CPU socket area reads above ambient environment, then any such software is obviously wrong and needs calibration. Even the 8W relative and 10W max TDP RD790 chip [under no load] and 1.1V doesn't idle less than 5 degrees below ambient, IC physics. Then I'll work my way up.

KTE: Hidden somewhere in post#1 is that info. MSR 0x19C contains the temp data for any Intel Core based processors. The DTS data is 7 bits long and is in bits[22..16]. Other info in that MSR are flags for thermal throttling which I'm hoping Real Temp is going to show correctly.

Ok, thanks. I'll check it out later, saves me the trouble of going through the docs.