Real Temp - New temp program for Intel Core processors

[jaredpace]

I have a T7100 core2duo and coretemp & realtemp. I keep hearing people want this tested on a mobile processor, is there anything i can do to help?

[Darth Vader]

did RealTemp shows subzero temps? how accurate?

[rge]

I have a T7100 core2duo and coretemp & realtemp. I keep hearing people want this tested on a mobile processor, is there anything i can do to help?

Yep, if you're willing. Find out what gradient exists at idle, from core to tcase temp at few different temps, like 50C, 60C, 70c tcase.

You would have to remove the heat sink, and place an accurate as possible temp monitor to back center of cpu, either thermocouple or accurate IR gun. If use IR gun, need to cover cpu with non-shiny thin tape, masking tape works ok, (black color apparently better if it exists), but no shiny tape.

Then position monitor so can see coretemp running and can read thermocouple/IR at same time. Record corresponding temps at same instance, using coretemp showing either delta to tjunction max or temps. (option, settings, check delta to tjunction if use that method)

Since mobile cpu has known tjmax, then would have better idea what temp gradient exists at idle from core to tcase.

Need to check at least couple temps in higher range, ie 70,80, since higher temps more accurate than lower.

Its difficult to get pics, but if second person is handy would be nice.

[day187]

Very nice thanks mayne

[jaredpace]

ah well i don't have a temp gun. didn't know that was a requirement as well.

[rge]

ah well i don't have a temp gun. didn't know that was a requirement as well.

Just out of curiosity what are your idle temps on T7100?, and at what vcore, mhz?

[jaredpace]

i can idle at 36 to 38C at 1.01 vcore. at 1.39 vcore it sits around 43C

at 1.8ghz w/ 9x multi at 1.39 vcore. at 1.8ghz 9x multi under load, vcore is 1.29 and after 20 mins of Prime, temps are 65C

these are temps reported by coretemp 96.1 and everest 1307

[rge]

Thanks.

[maatriks]

Any hopes of getting this working also on Vista x64? I would so so want it, but I also so so have x64..

[unclewebb]

Any hopes of getting this working also on Vista x64?

That is looking very promising now. I started switching over to a more modern driver last night so turning out a x64 edition looks simple now. Unfortunately a bad cold has slowed my a-s-s down today. x64 beta by bedtime today I hope.

KTE: I think loading a processor without a heatsink on it can give unpredictable results. That's why I gently took my cpus up to TjMax at the lowest possible voltage and MHz. I'm not sure but I have a feeling that the rate of temperature change might also effect processor shut down but I haven't found any Intel docs to back that up. It would be interesting if you could try my tests with an IR thermometer in one hand and a fire extinguisher in the other!

Darth Vader: As long as the DTS sensor in your processor are moving freely then sub zero readings shouldn't be a problem with Real Temp. I need to get outside for some real world testing of that but someone else posted a sub zero screen shot of my program earlier in this thread. When you run my Test Sensor feature what kind of numbers are reported? Are all 4 of your sensors moving freely and equally? They should.

[Ace-a-Rue]

sorry to hear about your bad cold; get well unclewebb...heavy dosaging of vitamin c helps to shorten the cold.

[lasershock]

That is looking very promising now. I started switching over to a more modern driver last night so turning out a x64 edition looks simple now. Unfortunately a bad cold has slowed my a-s-s down today. x64 beta by bedtime today I hope.

I'm looking forward to test the x64 version.

[clicker666]

I have a couple of questions, perhaps someone could enlighten me? From what I have been reading here, the TjMax may vary from processor to processor, is this correct? To allow for this you calibrate your CPU to determine the - -- 0 + ++ value that is most appropriate to your CPU at low power temp. (closest to ambient) How accurate could this method be? Wouldn't there be a lot of potential variables involved in determining the - -- 0 + ++ setting?

[davemelb]

Just tried this program. I have two questions.

1) Why is idle temperature the same whether I use cpu voltage of 1.1v or 1.5v? It is always the same at 22 on "0" degree calibration.

2) When I restart RT, Calibration resets back to "0". How do I get it to remember "++" after exit?

Thank you for your program.

[unclewebb]

... the TjMax may vary from processor to processor, is this correct?

There is nothing to prove that this is true or false at the moment. Personally I believe that TjMax is standardized across a process line so that all E8x00 dual core desktop processors with a CPUID of 0x10676 all use the same TjMax=95C. That assumption is what RealTemp is based on.

I think that if the factory is making 95C processors then each and every one is calibrated so that DTS=0 right at TjMax=95C. All of the 65nm mobile chips are documented to have a TjMax exactly equal to 100C and the new 45nm mobile chips are all set to exactly TjMax=105C so it doesn't make any sense to me that Intel would set all of the mobile chips to a fixed value but set all of the desktop chips to random TjMax values.

The (++) (--) feature is wide open to variables. For me, since the output from the DTS does not seem to be linear across the entire temperature range of the CPU, it provides me with the only way to make adjustments to idle temperatures to bring them more in line with measured temperatures as outlined in post#1. Without this adjustment, air cooled L2 processors either idle below ambient temperature which is impossible or you can use a TjMax=100C to cover up that problem and end up with them reporting that they are running far hotter than all of the other available Core2Duo processors. Neither solution is correct.

The Idle Calibration feature is just an approximation. It will never completely compensate for a DTS sensor that doesn't seem 100% accurate across the entire temperature operating range. It would be nice if a few users would try to calibrate as outlined in post #1 and report their findings. For me, my E6400 B2 needs a ++ calibration to keep it from reporting below ambient temperatures and my E8400 needs a -- calibration to bring it in line with my E6400. My CoreTemp testing which showed a 25C difference in idle temps between these two processors at the exact same core voltage, MHz and room temperature just doesn't make any sense. Real Temp idle readings should be within a degree or two now for these processors when calibrated.

My promise to have a x64 bit version ready by tonight is dead. I'm using Visual C++ 6.0 which doesn't support x64 without lots of downloading and screwing around. Without having access to a x64 OS, I'm not sure what I can do to make this happen. I was planning to make my code open source at some stage so if anyone can compile MFC to x64 then let me know. By switching over to the driver from
http://openlibsys.org/
it should be fairly easy for someone with the right compiler to create a x64 version. If there are no volunteers then x64 might have to wait for a week or two.

[unclewebb]

Why is idle temperature the same whether I use cpu voltage of 1.1v or 1.5v? It is always the same at 22 on "0" degree calibration.

That's usually a good sign that an on chip sensor is getting stuck. What does the Test Sensor button report? I need to add some code to save the settings.

[davemelb]

The Test Sensor reports "0" and "0" when I use 1.1v and 1.6Ghz. When I use 1.5v to calibrate, it reports 7 and 2.

I then make adjustment with "++".

[KTE]

KTE: I think loading a processor without a heatsink on it can give unpredictable results. That's why I gently took my cpus up to TjMax at the lowest possible voltage and MHz.

I did both incase what I wrote give you another impression. The load was placed later to see when and if it shuts down before blowing.

You have to keep in mind how power/temps work. TDP is what decides them and TDP varies from every CPU in the same batch. If Intel splashes a 95W TDP rating on a product, that doesn't mean it's 95W TDP. Many times it will be lower than 94W. Some even be 70W bins in the same SKU. That's why power will differ and so will heat/overclock. One CPU may handle more than another. There is no experimental evidence and most I've already experienced both sides of to believe that the throttling values are fixed on processors whilst users have experienced varying throttle initiation temps for a long time now a quick search of google can show. Like I stated, one CPU of mine is completely different to another. Let alone that, the case is the same with my Phenoms. One shut down 72C just within BIOS and one was all the way to 109C within Windows with the strong burning smell without throttling or shutdown.

The best any of us can say is, my experience with x CPU leads me to believe or tells me... because peoples findings will ultimately differ and this is about collective feedback and then the analysis.

I will test my Q6600 G0 in a bit as your first post mentions and with many more temperature tests. Just busy yet.

[rge]

Back in 1999 to 2004, there were many review articles written primarily at site linked to below discussing the same temp questions we are. But these reviewers were drilling holes in their cpus and placing accurate thermocouples on the cores, pulling off the IHS, measuring gradients, etc.

One article they used a thermocouple placed on the core of the cpu (drilled hole to it), comparing core temps to cpu casing temps. They described a "lag of 2C going from hot to cold because of resistance of CPU ceramic substrate", but only a difference of "0.4C...once the temp stabilized."
http://www.overclockers.com/tips443/

Edit: I think the above, and what I have seen in this thread, makes an average tjunction of 105, reported by coretemp, to be unlikely.

But regarding tjmax being a fixed number, note the quote "from intel specifically from one of our internal contacts for the processor documentation", that states, "Note Tj is not a fixed value and the IA32_TEMPERATURE_TARGET[15:8] value can vary from part to part. Tj is also not software readable."

http://softwarecommunity.intel.com/i...howThread.aspx
last 2 posts

[xploit]

Can we sticky this thread???

[unclewebb]

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

Just got finished uploading version 2.0. I decided to go up a notch in version numbers which represents the major change in device driver. The program is now self contained in one folder and doesn't require the IA32 driver anymore so you can remove it from your system by going into the Device Manager.

I located a Visual C++ .NET 2003 version. Does anyone know if I will be able to compile MFC code using this and can it output x64 code? I'll be looking into that subject tomorrow.

The above link is the same as before so no need to update if you have linked to this file on other sites.

[Qbical]

hey guys, ok so i havent been here in FOREVER so go easy on me lol
first off KILLER post uncle.

well basically i went out and got a shiney new E8400 in hopes to break 4ghz and stay there. seeing as i dont think i am too noob @ OCing i figured this should b fairly easy from what i heard. well needless to say after watcing my temps from beginning to end i was fairly confused. i then found this thread and had some suspicions confirmed (once again thanks)

as u can see below i was able to hit 4.3Ghz and had it stable enough that i was using it daily to game and do everything else for about 2 days.


today i decided to 'calibrate' my temps so i brought everything back down to stock


i was running at 4.3ghz and before i found RealTemp i was using CoreTemp.
here is a little scene from when i was gaming

20:38:18 03/03/08 4302.13 62 66
20:38:19 03/03/08 4302.13 62 67
20:38:20 03/03/08 4302.13 62 67
20:38:21 03/03/08 4302.13 62 68
20:38:22 03/03/08 4302.13 61 67
20:38:23 03/03/08 4302.13 62 67
20:38:24 03/03/08 4302.13 62 67
20:38:25 03/03/08 4302.13 62 68
20:38:26 03/03/08 4302.13 62 68
20:38:27 03/03/08 4302.13 61 67
20:38:28 03/03/08 4302.13 62 68
20:38:29 03/03/08 4302.13 62 68
20:38:30 03/03/08 4302.13 62 67
20:38:31 03/03/08 4302.13 62 68
20:38:32 03/03/08 4302.13 62 67
20:38:33 03/03/08 4302.13 62 67
20:38:34 03/03/08 4302.13 62 68
20:38:35 03/03/08 4302.13 62 67
20:38:36 03/03/08 4302.13 62 67
20:38:37 03/03/08 4302.13 62 67
20:38:38 03/03/08 4302.13 62 67
20:38:39 03/03/08 4302.13 62 67
20:38:40 03/03/08 4302.13 62 67
20:38:41 03/03/08 4302.13 62 67
20:38:42 03/03/08 4302.13 62 68
20:38:43 03/03/08 4302.13 62 68

here is no load
19:46:06 03/04/08 3005.61 41 54
19:46:07 03/04/08 3005.61 40 54
19:46:08 03/04/08 3005.61 40 51
19:46:09 03/04/08 3005.61 40 51
19:46:10 03/04/08 3005.61 40 52
19:46:11 03/04/08 3005.61 40 51
19:46:12 03/04/08 3005.61 40 51
19:46:13 03/04/08 3005.61 40 51
19:46:14 03/04/08 3005.61 40 51
19:46:15 03/04/08 3005.61 40 52
19:46:16 03/04/08 3005.61 40 52
19:46:17 03/04/08 3005.61 42 53
19:46:18 03/04/08 3005.61 41 52
19:46:19 03/04/08 3005.61 40 51
19:46:20 03/04/08 3005.61 41 52
19:46:21 03/04/08 3005.61 40 51
19:46:22 03/04/08 3005.61 40 51
19:46:23 03/04/08 3005.61 40 51
19:46:24 03/04/08 3005.61 40 51
19:46:25 03/04/08 3005.61 40 51
19:46:26 03/04/08 3005.61 40 51
19:46:27 03/04/08 3005.61 40 51

there are several ?'s i have from all the stuff i have been doing and seeing here.

1. my Core#0 is ALWAYS 10 degrees less then my Core#1 ,why?
i was reading about the thermal sensor being stuck, but mine still moves but its just always ~10d less.

2. i noticed that my CPU has no revision #, should i be concerned about that?
According to Coretemp my revision is C0; so it says.

3. y does the test say both my cores are 5 but i KNOW my core 0 is always lower then 1. so wouldnt the test read like 1 and 5 or something like that?

4.idk what to believe anymore. my room ambient temp is about 26-27C and maybe even lower at night. right now RealTemp shows (27C/38C) but a few hours ago it was at (31C/41C) all this is stock no OC. that is only a few degrees off of what the room actually is, so does that mean i should use the (0) setting? i mean my case has great airflow and when i shoot my temp gun around everything seems nice and chilly. i mean my fricking 8800GTS is @ like 47C idle.
i even added a new mini fan to blow across the ram and CPU so i have a hard time believing a 70C CPU temp

i would love any kind of suggestion other then me just cookin this thing. lol
btw i have the CPU and NB on 1 cooling loop w/ a DD dual 120mm rad. that has both fans on it. see below

ASUS P5K Deluxe
C2D E8400 @ stock for now
water cooled CPU and NB-dedicated loop
2GB Corsair Dominator DDR2 @1600mhz 5-5-5-15 2T
EVGA 8800GTS BIOS V flash (660/840)
watercooled-dedicated loop
2x 200GB Maxtor SATA
SB XFI Platinum
PC Power & Cooling Silencer 610 PSU
Dell 2407WFP

[hersounds]

so,unclewebb...in my picture..who say the true?



of course!...is important that update to realtemp 2.0 ?

[t_ski]

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

The link to the new version in your last post is wrong. It should be this (from the first post):

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

The last link has "..." in the middle of it, like you copied the text and not the link.

[unclewebb]

Thanks t_ski. It might have something to do with the meds I'm on today! My previous post has now been corrected.

hersounds: I like 29C much better than 39C. Updating to 2.0 will let you remove the IA32 driver from your system. As a user I'd update for that reason.

Qbical: It is obvious that there is a problem with the DTS sensors in the E8400 that you are using. The TestSensors test shows that they are both moving freely and equally but being 10C different from each other shows that Intel quality control screwed up. Keep overclocking your processor. When it blows up, you'll have a good reason to return it. I think your lower temperature reading in Real Temp is probably close to the correct temperature but when you are having sensor issues it's impossible to say unless you pop off your heatsink and point an IR gun at it.