Real Temp - New temp program for Intel Core processors

[unclewebb]

RejZoR: CPU usage spikes when you're using the new VistaDelay feature in RealTemp is not RealTemp hogging your CPU. RealTemp is using zero CPU cycles while it's waiting for the rest of the Vista baggage to load up. It's Vista or other software that is overloading your E4300 at start up. Open up the Task Manager and you'll probably see what is guilty.

If the calibration of your E4300 was at 0.85 volts then you didn't follow my recommendations. You made me write out all those steps and by step 2 you had already decided to do things your own way.

2) Set your core voltage to approximately 1.10 volts.

The reason I recommend this voltage is so a fair, direct comparison can be made between the testing with an IR thermometer I've done and the testing that you're doing.

Anyhow, 65nm sensors having an error of 7C or 8C at idle is completely normal. With less voltage the error is likely greater but I haven't done any testing to back up that theory. Calibrating based on the accuracy of an outdoor thermometer is not very scientific. Here's a pretty thorough test I did on an E2160 which is very similar to your E4300.

http://www.xtremesystems.org/forums/...&postcount=423

The best thing about RealTemp is that if you think all of my testing is bunk then you can choose whatever TjMax you like. My opinion, based on my testing, is that your choice of TjMax=95C for your E4300 might make your idle temps look nice but your load temps are now being reported 10C too high.

Let me know how much time delay is needed so RealTemp can Start Minimized for Vista users that are using UAC.

[WoZZeR999]

Personally, I like the old setup of RT, but I really only use mini-mode. The old version is a little cluttered, but it makes efficient use of screen real estate. I don't really care to always know what my min and max temp are, so maybe a setting to show hide would help. When using for day-to-day stuff, knowing what the min and max temp where doesn't really help me. I don't really need any of the aux information other than maybe freq/multi. ProcHot would be helpful with time/date on it as a pop-up box.

That new setup is going to that CoreTemp look/feel, which is a bit to ...industrial... for me.

[loonym]

Personally, I like the old setup of RT, but I really only use mini-mode. The old version is a little cluttered, but it makes efficient use of screen real estate. I don't really care to always know what my min and max temp are, so maybe a setting to show hide would help. When using for day-to-day stuff, knowing what the min and max temp where doesn't really help me. I don't really need any of the aux information other than maybe freq/multi. ProcHot would be helpful with time/date on it as a pop-up box.

That new setup is going to that CoreTemp look/feel, which is a bit to ...industrial... for me.

Well said WoZZeR unc's got everyone playing follow the leader now. If anyone seriously thinks any of these other applets would have implemented the changes they have recently without this kind of motivation think again. It would have simply been a regurgitated 'more of the same'.

[RejZoR]

This delay thing is weird. Soemtimes it works, sometimes it doesn't. 30 sec delay. If i use delay in Task Sceduler it always launched properly. Hm...
Also this calibration is like gambling. I've tried with 1,1V and it was just a +4°C difference. So i'm not even gonna bother anymore. Whatever it shows i'll take that as true...
Who knows what's actually right and what isn't when everyone get everything different...

[Unrealer]

OT: what vcore (on full) do u need to do 3.4Ghz? and what temps do you get on full with core temp?

[RejZoR]

I was testing on 1,5V but i've lowered it to 1,48750V. It's not Orthos stable but appears to work just fine in games (tested Tiberium Wars so far). At 388x9 it's not stable anymore and restarts when it should login to Windows. My CPU has a very low FSB wall so pumping more volts does pretty much nothing except more heat. But hey, even 3,4GHz is not all that bad. Max temp was 62°C according to RT (while playing Tiberium Wars).

[unclewebb]

RejZoR: Intel did not design or calibrate their on chip digital thermal sensors to report accurate idle temperatures. If you closely follow the calibration procedure I outlined, your reported temps will be very accurate from idle to TjMax without having to invest a lot of money in calibration equipment. Something a little more accurate than your back yard thermometer would be a good place to start! If you had your own IR thermometer you might get a degree closer to the actual core temperature but it's not worth the investment for this.

Thanks for your testing of the Start Minimized bug in RealTemp when using Vista. At least I know what's going on now. I'll see if I can come up with a different work around that is compatible with Vista.

I might introduce a second GUI version of RealTemp at some time to give users another choice. Any design ideas are appreciated.

[SiGfever]

RejZoR: Intel did not design or calibrate their on chip digital thermal sensors to report accurate idle temperatures. If you closely follow the calibration procedure I outlined, your reported temps will be very accurate from idle to TjMax without having to invest a lot of money in calibration equipment. Something a little more accurate than your back yard thermometer would be a good place to start! If you had your own IR thermometer you might get a degree closer to the actual core temperature but it's not worth the investment for this.

Thanks for your testing of the Start Minimized bug in RealTemp when using Vista. At least I know what's going on now. I'll see if I can come up with a different work around that is compatible with Vista.

I might introduce a second GUI version of RealTemp at some time to give users another choice. Any design ideas are appreciated.

Yeah, they got to be what +/- 5c?

[pitadotcom]

Hm, it appears Tj.Max 85 is not the correct value for my E4300. However 95 appears to be. I've totally downclocked and undervolted the core to 1,2GHz @ 0,85V, all fans max and case opened, classic outdoor thermometer was placed near computer for temperature verification.

yeah i got mine on 95 as well. its just looks more realistic when doing the idle temps test.

[rge]

AT RejZor, pitadotcom, and others using tjmax to calibrate idle. Unclewebb and computronix have already answered this many times, so perhaps an illustrated version using unclewebbs graph.

First, accurate idle temps is better evidence of a wrong tjmax than a right one, since most DTS have nonlinear, uncalibrated idle temps. Unclewebb's testing, others testing/confirming, and Intel has confirmed via anandtech, and also by their documents that temps are nonlinear, accurate at load 20-30 from tjmax, but inaccurate at idle.

If you believe intel and the testing, the pic/graph illustrates the result of adjusting tjmax to calibrate inaccurate idle temps....

The Black line represents the true temperature if a DTS was calibrated properly at high and low end and was linear.

The Red line is the actual DTS nonlinear sensor output of E4300.

Raising tjmax, going from red dot/red line to the blue dot/blue line, you lift that entire curve, so at idle you are now accurate, but you simply transferred the error to load. (What you should have done was left tjmax alone, and just calibrated the low inaccurate end with RT calibration, then you would have ended up with the red line merging completely with the black line.)

To get accurate tjmax's, you must use IR testing in the accurate sensor range, the high end. You can not conclude anything by looking at inaccurate idle temps.

If you do use tjmax to adjust idle, then you are claiming DTS output is linear, despite intel stating otherwise, and despite testing showing it is not.

[Mr Roboto]

@unclewebb

Comparing RealTemp, CoreTemp and Everest, RealTemp always reads about 10c lower, obviously because of the default 95c Junction Temperature. My Asus AISuite Temp readings and RealTemp are the same though (44c Idle 51c Load).

However in Everest's options preferences\stability checking the box "Measure CPU Temperature Using ACPI" gives me the same 44c. CoreTemp is the only program reading temps higher then. Any idea why such a difference when changing the measurement type? Thanks.

[unclewebb]

rge: Thanks for the modified graph. I've been meaning to show that graphically to explain what people are accomplishing by adjusting TjMax when they probably shouldn't be. The graph shifts upward, you create some realistic looking idle temps but you've created inaccuracies in your reported temps everywhere else between idle and TjMax. Personally I wouldn't adjust TjMax one degree unless I had an IR thermometer or similar device to verify that the TjMax RealTemp is using is wrong.

I keep telling people that these sensors are not accurate at idle and then they keep using inaccurate information to decide whether RealTemp is using the correct TjMax or not. I'll never be able to convince everybody but at least I've been able to convince a lot of people that their software is not being honest and that RealTemp is about the best choice for this problem.

Mr Roboto: I don't quite understand your question but here's some info that might help.

Reported Temps = TjMax - Digital Thermal Sensor Reading

RealTemp uses TjMax=95C and CoreTemp uses TjMax=105C. If you plug those numbers into the formula then where the 10C difference in reported temperatures is coming from is pretty obvious. Both programs read the same Intel sensor and agree on that but interpret the data differently. I decided on 95C based on my IR thermometer testing. I'm assuming that CoreTemp went with 105C based on Intel documentation for the mobile processors which I don't believe is relevant to the Intel Desktop processors.

I'm not 100% sure what Everest or Asus AISuite are reading but I assume both of them are reading the CPU temp sensor in the center of the chip. The calibration for this sensor isn't always accurate depending on the bios so it's nice to see that it seems to work on your board and that it's showing the same as RealTemp. In theory, during quick load transitions, the readings from RealTemp might be more accurate than your other CPU readings but these 45nm chips are getting so small that the sensors are very close together and if they are all calibrated and working properly, you might not see much difference between them. If that doesn't quite answer things for you then just try again.

[Mr Roboto]

Sorry to make that so confusing. You answered my question. Thanks for a great little app.

Also like people stated many times these sensors are terribly inaccurate at idle. Sometimes my BIOS as well as AISuite reads 124c. Obviously that's impossible. Once you give it a little juice it comes right down.

[RejZoR]

Hm, but wouldn't CoreTemp work right if you change the Tj.Max value to the correct one (85 instead default 100 for E4300)? It does offer the option to change it... Or is there some other calculation involved other than only Tj.Max value ?

[unclewebb]

Might as well be calling this 2.70 RC1 because it's about ready for release I hope as the next major version.

http://www.fileden.com/files/2008/3/...alTempBeta.zip
Download, unzip and copy RealTemp.exe into your RealTemp folder if you want to check it out.

I tried to do a work around to get Start Minimized working for users that use UAC. If Start Minimized was not working properly for you before then give this version a try and see if it is any better. Hopefully I didn't screw anything up.

I also added a sound file option for the alarm feature. The file is called RTWarning.wav and is included in the above download. If you don't like this sound then blame CompuTronix! Better yet, you can use whatever sound you like as long as you rename it RTWarning.wav and stick it in the RealTemp directory. Keep the sound file under 5 seconds in length or it might cause some trouble. I haven't checked this out yet or built in any safety features for those that like to experiment. Trying to play a 5 minute wav file for your warning probably wouldn't be a good idea right now.

RejZoR: At full load, CoreTemp or Everest work fine if you adjust TjMax to 85C. The only problem is that other programs do not give you any way to compensate for thermal sensors that have been proven to be inaccurate at reporting low temperatures. The amount of error in these sensors can approach 10C at very low idle temperatures which is why some people just bump up TjMax by 10C but we all know now that changing TjMax won't cure where this problem originates from.

[Richard Dower]

Keep up the sterling work Uncle!

[pitadotcom]

oops. im actually on tjmax 85 and +2 calibration. my bad. from my understanding i thought that the +2 calibration would make my tjmax 95. i guess i was wrong.

[RejZoR]

unclewebb, change the Ok and Use buttons as i've suggested some time ago. Also change the Use button not to close the window (plus rename it to Apply). "Ok" should remain as "Ok" and should save the settings and close the window.

[RealTelstar]

I like the "save" in place of "ok", but "use" should definitively be changed to "apply"

[RejZoR]

Start minimized seems to work fine now. Will test it some more but it looks like it's ok.

[CBird28]

sorry if this isnt needed but how comes its reading that much lower than coretemp and everest?

with a room temp of 26c

everest : 53/53
coretemp: 52/52

real temp: 42/42

i dont trust realtemp to be honest, considering its reporting completly different tests to everything else

[RejZoR]

Hm, what about integrated table of Tj.Max values depending on detected CPU ? For example if RT detects E4300 like in my case it would automatically set the Tj.Max to 85.
For some other CPU it would be 95 and for some other 100 etc etc. Only problem i see for doing this is gathering correct Tj.Max values. Applying them according to CPU name should be easy.

[WoZZeR999]

sorry if this isnt needed but how comes its reading that much lower than coretemp and everest?

with a room temp of 26c

everest : 53/53
coretemp: 52/52

real temp: 42/42

i dont trust realtemp to be honest, considering its reporting completly different tests to everything else

Read the first page, that will explain why it's a different temp. I trust realtemp because there is real world (read: actually using a thermometer on the processor IHS) testing.

[unclewebb]

i don't trust RealTemp to be honest, considering its reporting completely different tests to everything else

The competition does absolutely zero real world testing and basis their results on non-existent documentation from Intel and you choose to trust them over RealTemp. That's kind of funny.

What isn't funny is how far off CoreTemp is when reporting the temperature of my E8400. Here's some of the testing that I've done. Ask the competition where their testing is and what they're basing their results on.

http://www.xtremesystems.org/forums/...postcount=1599

pitadotcom: oops. im actually on tjmax 85 and +2 calibration. my bad. from my understanding i thought that the +2 calibration would make my tjmax 95. i guess i was wrong.

Sounds like your calibration is just right! Most of the 65nm sensors seem to read too low. My original E6400 with a TjMax=85C also needs a calibration offset of +2 for very accurate temps from idle to TjMax.

RejZoR: Hm, what about integrated table of Tj.Max values depending on detected CPU ? For example if RT detects E4300 like in my case it would automatically set the Tj.Max to 85.

That's what RealTemp is doing right now. It automatically detects your E4300 and sets it to TjMax=85C and sets it to TjMax=95C for my E8400. If a user decides to select a different TjMax in the Settings window then it will go along with that number. I was going to publish a list of TjMax values that RealTemp uses. Unfortunately, too many users have preconceived notions about TjMax. They pass judgment on a program without reading the documentation or doing any testing so I said to hell with it.

[unclewebb]

I think Save and Use come from my old Amiga programming days. Save makes more sense to me than OK but OK seems to be the generally accepted Windows term so I'll be using OK.

As for Apply, that's another Windows term that I guess I should be using. Should the Settings window stay opened after you click on Apply? That's what RejZoR thinks should happen. A lot of the functionality in the Settings window is based on any new settings being automatically applied. I find this gives direct feedback as soon as you enter in a new calibration value and Tab over to the next box. Any thoughts?

I'm just cleaning up any loose ends before the next release. Very happy to hear that the Start Minimized issue with Vista might finally be solved.