Real Temp - New temp program for Intel Core processors

[cadaveca]

unclewebb, i take high temps from CoreTemp as realtemp reports sub-ambient temps @ idle. I prefer your program, however, I cannot accept 10C sub-ambient reporting.

But i will be clear, this problem is not yours..cpu sensor is not accurate.

[loonym]

unclewebb, i take high temps from CoreTemp as realtemp reports sub-ambient temps @ idle. I prefer your program, however, I cannot accept 10C sub-ambient reporting.

But i will be clear, this problem is not yours..cpu sensor is not accurate.

Why not just calibrate the ini file in realtemp?

[cadaveca]

Because then it reads the same temps as Coretemp.

Coretemp ahs teh same problem at lower speeds/volts though...lower than ambient reported.

[unclewebb]

cadaveca: My E8400 when uncorrected at idle will report temperatures that are 8C too high in my cool basement but is dead on from 60C to 95C. It seems like a pretty average chip and I know there has to be ones out there that are both better and worse than that.

Everyone needs to accept the fact that the digital thermal sensors weren't designed or calibrated by Intel to provide accurate idle temperatures and the amount of error when you are a long ways from TjMax is significant. This isn't just a new 45nm issue either. It effects all Core based processors. My original B2 - E6400 also has an error of 7C or 8C. My E8400 reads too high and the E6400 reads too low like many other B2 processors. The only difference between 45nm and 65nm is that the 45nm chips are having sticking sensor issues in the temperature range where many people use them. There have been reports of the 65nm series also getting stuck but it seemed to happen at a lot colder temperatures so hardly anyone ever noticed.

I worked with a couple of users that had sensors that produced data along two totally different slopes but I was still able to come up with a unique correction factor to get their readings a lot closer to reality. RealTemp can't make every processor perfect but as long as your sensors aren't stuck, it can make a big difference and get your readings much closer to the actual temperature. Even really bunged up sensors still seem to move quite linearly even if they aren't moving along the line that they are supposed to be on.

If you would like me to work with you then just send me a PM with a few details and then I can tell you what data I need.

To see two bizarre corrections that were needed, go into the INI file and type in User=1 or User=2. It will really mess up your reported temps when you use one of these calibration curves on a normal processor.

[cadaveca]

I totally agree with you about the sensors 100%. I feel the same way about software readings for voltages...variations in process affect accuracy. I wonder how much extra work this has created for Intel!


I must say thanks very much for the offer of your time...that's mighty awesome of you!

In the end, temps aren't that important to me, so I'll save your time for someone who really needs it...I know my cpu is not overheating, and my cooling is up to snuff...

I am in no alluding or hint at any issues with realtemp...I use both it and coretemp(obviously, to have made comparisons). The problem is my cpu, but in the end, it's not that big of an issue for me. I'm more interested in the variation from idle to load, and not exact readings...they just can't be had easily in some instances!

[unclewebb]

The only voltages I trust are ones that are displayed on my digital volt meter. The ones coming from software that reads iffy data are never as accurate.

My offer to help any user with really messed up sensors stands. As long as the sensors aren't sticking, you can significantly improve the accuracy with a tweak or two.

In the end, temps aren't that important to me

I agree with you 100%. When the 45nm series came out I almost said, "Why bother writing another temperature utility program?" For 24/7 use, even at 4 GHz, my E8400 runs so cool that temps really aren't important. With CoreTemp grossly exaggerating my E8400 temps and then the sticking sensor issue showed up so I said, "What the hell and this project has kind of grown from there."

[Milamber]

Has the core temp author been in here and say "G'day" or been in contact?

[unclewebb]

By tweak I mean actually using the Idle calibration factors to get one's idle temperatures a little closer to the real temperature. Most users don't seem to bother with this and if you are only interested in your load temps and your load temps are at least 60C then I guess it's not that important. If your processor is really out to lunch and can't be fixed up with these simple adjustments then you can ask me to have a look at your data and I will try to create a custom calibration curve for you. This is rarely necessary but is a nice option to have.

Has the core temp author been in here and say "G'day" or been in contact?

I'm sure he's been quietly around here once or twice.

I check out CoreTemp. He has some ideas I like and competition is good for both programs. He's been making a few changes lately and so have I.

I like CoreTemp's ability to show the temps of all 4 cores in the System Tray so I've taken that one step further. A user will get to choose exactly what core or cores he would like to see in the system tray. With a Quad you might only want to see one of the first two cores and one of the second pair of cores. It will be easy to configure so a user can see just what he wants. I've only got two colors so far but you could do something like set one pair of cores to black and the other pair to white. I like software that you can set up however you like.

I'm presently working on the option to show Minimum, Maximum and maybe even the average core temp in the Tray since the last Reset.

I've also simplified entering TjMax adjustment values. In the next version you will be able to simply enter in TjMax0=97 or whatever number you like. I still don't believe most users need to adjust TjMax but I decided to make it simpler for those that do.

I'll also be including one digit after the decimal point for the Idle factors. Sometimes you don't know if -2 or -3 would be best so you'll be able to split the difference and use -2.5 or -2.3 or -2.7 or whatever.

The best way to find out what's appropriate for your processor is to do the 1600 MHz / 1.10 volt test and compare your reported temps to your room temperature or water temperature. This was the simplest thing I could come up with and it seems to work good. It's explained more thoroughly in the docs.

The last thing that needs to be looked at is trying to get the MHz display 100% as accurate as CPU-Z. It works great on my computers but my simple solution is not yet perfect for all users. The problem is that Windows doesn't come with any accurate way to measure elapsed time. You sometimes have to use multiple timers and compare the results to try and find which one is telling the truth. I like starting simple and only making it as complicated as it needs to be to get the job done. Less coding that way!

[Unrealer]

New mobo, Heatsink & fan...

DFI LanParty P35 Dark, Thermalright 120 Ultra Extreme & Noctua P12



Ambient temp 24.5º
Noctua @ MAX (1350rpms)
TRUE & E4300 lapped

IDLE:



EDIT:

65 (core temp) / 50º (real temp) on FULL with OCCT .........

[rge]

Ambient temp 24.5º
Noctua @ MAX (1350rpms)
TRUE & E4300 lapped


50 (core temp) / 65º (real temp) on FULL with OCCT .........

Typo?... thats backwards, it would be 50C load with real temp, and 65C load with core temp.

But that one is easy, with EIST enabled at idle and low volts with good cooling like you have, you would expect core temperature to be just above ambient like real temp reports using tjmax 85, and not 16C above ambient as core temp reports using tjmax 100.

I think that is simply more evidence that tjmax is 85....meaning your correct load temp is closer to 50C (within the usual DTS error of reporting at that range)

And your bios did the typically crappy job of calibrating the diode for cpu temp...as it is reporting 3.5C below ambient.

[Unrealer]

Typo?... thats backwards, it would be 50C load with real temp, and 65C load with core temp.

Yeah, sorry

But that one is easy, with EIST enabled at idle and low volts with good cooling like you have, you would expect core temperature to be just above ambient like real temp reports using tjmax 85, and not 16C above ambient as core temp reports using tjmax 100.

I must try that, cause i don´t know what programa is lying.. ¬¬

I think that is simply more evidence that tjmax is 85....meaning your correct load temp is closer to 50C (within the usual DTS error of reporting at that range)

So, it´s 50C it´s the real temp, i think is good...

Anyone with a 4300??


EDIT: the same full temp with a 120CFM FAN

[rge]

EDIT: the same full temp with a 120CFM FAN

Download speedfan and make sure your 120CFM fan is actually going any faster. I am assuming you did not plug this fan into cpu fan slot, where it will spin slower until temps get much higher.

But, since both programs accurately read DTS sensor output, if load temp was measured the same with faster fan, the really funny thing is the temp WAS THE SAME.

[Unrealer]

Download speedfan and make sure your 120CFM fan is actually going any faster. I am assuming you did not plug this fan into cpu fan slot, where it will spin slower until temps get much higher.

i plugged it to the PSU... is spinnings @ max RPMs.

EDIT:

IDLE 1.2GHz 1.2v

Core Temp: 40ºC
Real Temp: 25ºC



FULL 1.8GHz 1.3v

Core Temp: 50ºC
Real Temp: 35ºC






****Ambient temp : 24.5º
****TRUE with a 120CFM FAN

[rge]

With fan spinning at max on your setup, at 1.2Ghz 6x200, and 1.2V, your core temp should be closer to ambient, not 15C above. So again, looks like real temp with tjmax of 85 is going to give you the most accurate temps. No way is <6W TDP going to idle 15C above ambient with that cooling.

[SiGfever]

Hello,

Here are my idle temps on my Q6600 @3.62Ghz 1.32v (CPUZ)

Core Temp...35,35,32,33
Real Temp...30,30,27,28
Everest Ultimate...35,35,32,33

Everest and Core Temp report the same and your program shows the temps much lower?

[unclewebb]

SiGfever: Read through the documentation for RealTemp so you can learn a little more about what RealTemp is all about.

Reported Temperature = TjMax - DTS

DTS is the reading from the digital thermal sensors which all programs don't have a problem reading. Everest and CoreTemp assume TjMax=100C which seems to be based on some old, irrelevant and non-existent Intel documentation for their mobile processors. RealTemp uses TjMax=95C for your GO Q6600 because I actually have a G0 Q6600 and the first day I got it I ran it without a heatsink and measured the core temperature with an IR thermometer and discovered that TjMax=100C was not true.

If you don't have any faith in my testing then head into the RealTemp.ini file and set this:

TjMax0=1
TjMax1=1
TjMax2=1
TjMax3=1

and then RealTemp and those other two programs will report the exact same thing.

Unrealer: Thanks for testing the E4300. The original version of CoreTemp used TjMax=85C for the E4300 and then users complained about sub-ambient temperatures due to the inaccuracies in these sensors at idle so the programmer caved in and boosted it up to 100C to keep everyone happy and quiet. I don't think TjMax=100C applies to any desktop processors but back then it was like flipping a coin. If 85C isn't right then we'll make it 100C. After that, every other program followed along with out doing any proper testing, until RealTemp came along that is.

50º (real temp) on FULL with OCCT .........

Thanks for your vote of confidence before doing any testing.

[SiGfever]

SiGfever: Read through the documentation for RealTemp so you can learn a little more about what RealTemp is all about.

Reported Temperature = TjMax - DTS

DTS is the reading from the digital thermal sensors which all programs don't have a problem reading. Everest and CoreTemp assume TjMax=100C which seems to be based on some old, irrelevant and non-existent Intel documentation for their mobile processors. RealTemp uses TjMax=95C for your GO Q6600 because I actually have a G0 Q6600 and the first day I got it I ran it without a heatsink and measured the core temperature with an IR thermometer and discovered that TjMax=100C was not true.

If you don't have any faith in my testing then head into the RealTemp.ini file and set this:

TjMax0=1
TjMax1=1
TjMax2=1
TjMax3=1

and then RealTemp and those other two programs will report the exact same thing.

Unrealer: Thanks for testing the E4300. The original version of CoreTemp used TjMax=85C for the E4300 and then users complained about sub-ambient temperatures due to the inaccuracies in these sensors at idle so the programmer caved in and boosted it up to 100C to keep everyone happy and quiet. I don't think TjMax=100C applies to any desktop processors but back then it was like flipping a coin. If 85C isn't right then we'll make it 100C. After that, every other program followed along with out doing any proper testing, until RealTemp came along that is.



Thanks for your vote of confidence before doing any testing.

I was not saying I did not have faith in your program, I just wanted to understand the different readings. I like your temps much better.

Thanks for taking the time to address my question. Also what version allows the temps to be displayed in the Systray?
John

[unclewebb]

SiGfever: The version that shows temps in the System Tray should be available sometime in the next few days.

[Demo]

Just noticed that 2.5 is out, thanks for the update. keep up the good work uncle

[unclewebb]

Just noticed that 2.5 is out, thanks for the update. keep up the good work uncle

You're falling behind here. I'm almost ready to release 2.6 with lots of additions and new features like temps in the tray area, finally. In house testing is looking good. Maybe there will be a beta for the XS guys this weekend. Anyone still hanging around here after 40 pages deserves something.

[Demo]

I've given time to read all the posts ive missed for the 30days that ive been out of town.. this thread is extremely informative to me even though i dont have a 45nm CPU but its a good read for stock knowledge.

[kpo6969]

This is the most informative thread I've ever read. Thank You

[emoners]

This is the most informative thread I've ever read. Thank You

true...not only has it given info about RealTemp per se, but also about DTS & stuff. +1 for unclewebb!

[zlojack]

unclewebb, I've got a question for you.

Last night I did the test of lowering my CPU to 1.2MHz and my vcore to 1.2v. My room temp was about 19*C and I was getting idle temps across the four cores of about 30*-34*C

Is this normal or should I perhaps do some calibration? I have been worried about high temps on my CPU which idles in the low 40s and can get up to the mid to high 60s under load. I have had high temps all along and I am wondering if it could be an issue with my motherboard as I am getting much lower temps on my secondary rig than on my main rig.

The secondary rig is a P5K Vanilla with a TRUE on a Q6700 and under full load at 3.2GHz and 1.4v(bios) 1.25v load in CPU-Z I'm getting into the low 40s!!!

I am probably going to try my QX9650 on that board and see how it does for temps, but do you think my mobo could be either a) overheating my cpu's or b) screwing up the temp reporting?

[unclewebb]

zlojack: Always nice if you can post a screen shot when you do a test so I have some numbers to look at.

I'll start by assuming that you are not presently using an Idle calibration factor. My testing was at 1.08 volts (core) so at 1.20 volts, my numbers aren't directly comparable to yours.

Your choice of motherboard shouldn't really effect your core temperatures as the DTS sensors are within the CPU and will read the same no matter what motherboard you put it in as long as everything else like voltage and cooling are the same.

During testing I also have an open case to reduce other heat sources including a hot graphics card or two.

Are you using the same cooler on both? If so then I would suggest an Idle setting of about -1 or -2. When the new version of RealTemp comes out then you could be a good candidate for Idle=-1.5

I managed to sneak a couple of more features in while the boss wasn't looking.



In the tray you can choose whether you want to look at the Minimum, the Maximum, the current Core Temperature or even the Average temperature since starting the program or hitting the Reset button. Does the competition have any features like that? I don't think so!