Need Help with RealTemp 3.42

[bushisland]

http://fohttp://forums.techpowerup....ent.php?attachmentid=31079&stc=1&d=1259712265
This was done with case on side; case panel off. Room temp near case 25C. Case temp inside case(still air) 26C. Holding probe just above CPU plate cover 32C.Touching top of heatsink base 32C. I am setting Core voltage to 1.13125V to get close to 1.10V in CPU-Z, because of VDroop. See attachments one and two
I also set Core voltage to 10.0V and allowed Vdroop, to see if it makes a different. See attachment three.
This unit I am using was just calibrated 2 months ago, so the temp readings should be OK.
On both voltage settings, all senors are moving. when the load % settles down(after opening RealTemp). Guessing, I would say the core1 set at 103 TJMax might not change as frequently as the rest.

 

[unclewebb]

My best guess is that your sensors read 4C or 5C too high at idle. I think the sensors on my E8400 also read too high, maybe 3C would be my guess. Once I come up with a better calibration formula then if you want you can use some negative calibration factors to drop your idle temps. To be honest, most users don't bother. Most people know that these sensors were not designed for accurate temperature reporting and just accept what ever numbers they show.

I was working on an obscure bug in the RealTemp / RivaTuner plugin today. Once that is done, I'll add an extra variable to the calibration formula to help out users that are interested in calibrating these things. After calibration, I think they're fairly accurate but it's difficult to prove without some pricey testing equipment.

 

[burebista]

I was working on an obscure bug in the RealTemp / RivaTuner plugin today.

A-ha, so now I'm an obscure bug.
/joke

Good job Kevin for finding and fixing that annoying bug.

 

[bushisland]

unclewebb,
Thank You very much. I'll be waiting.
bushisland

 

[unclewebb]

http://www.sendspace.com/file/5100no

Don't worry bushisland, I didn't forget about you.

I think we settled on something like this for accurate load temperatures.
TJMax = 99, 103, 105, 102

When I did my original testing, I found that there was virtually no slope error in these sensors from about 70C to 100C. Unfortunately, I changed my calibration formula based on the graphs that Intel presented at their developers forum last year. In hindsight, I think I took the information they presented way too literally. When Intel decided to come clean about these sensors, I was expecting to see some engineering data but the IDF forum was more just a general conversation and they didn't get into the specifics.

You can add this to your INI file so that the calibration formula has no effect from 70C to 100C.

OldFormula=1

By using this and the above TJMax values, you should be able to get them to line up at full load and they should be able to line up better at idle too. It probably won't be perfect but it should be more flexible than before.

Give it a try and post some results.

Edit: You'll need to use larger calibration factors when combined with this new INI file option.

 

[bushisland]

I down loaded version 3.50 and removed the semi-colon in front of OldFormula=1. I set TJMax loads: core 0-99, 1-103, 2-105 and 3-102. I then ran a load and low idle test. I did not touch idle calibration settings, or should I have?
httphttp://forums.techpowerup.com/attachment.php?attachmentid=31217&stc=1&d=1260129522://forums.techpowerup.com/attachment.php?attachmentid=31216&stc=1&d=1260129522

 

[unclewebb]

The new INI key, OldFormula=1 just changes the formula used. You'll still need to come up with your own calibration factors to balance your temperatures at idle.

The original formula should let you lower your idle temperatures on your last core without screwing up your full load temperatures.

An idle factor somewhere around -5.0 should drop the last core about 3C.

 

[bushisland]

I have adjusted the last core as advised. Two screen shots attached. One under full load and the other at idle. Now if you believe my temps are reading 4-5 degrees higher at idle and I gave a minus factor on each core to compensate, would that screw something else up. I guess I want a perfect world and there just isn't one.
http:/http://forums.techpowerup.com/attachment.php?attachmentid=31219&stc=1&d=1260135178/forums.techpowerup.com/attachment.php?attachmentid=31218&stc=1&d=1260135178

 

[unclewebb]

That's why I don't bother trying to calibrate these things unless they are way out of whack.

As a perfectionist, these sensors are too imperfect. Trying to get 100% accurate temperatures out of them from idle to TJMax is simply not possible. They'll only end up driving you nuts.

With the new formula, you can use some negative calibration factors to bring your reported idle temperatures down a few degrees and try to line them up better. The formula is linear so if you use 4C correction at idle then there will be 2C of correction at about 55C and 0C of correction at 70C and beyond.

Because these sensors only output data in whole numbers, they seem to vary a little from one day to the next so a perfect calibration one day might not be as perfect a month or two from now when your room temperature changes or the sun comes out or .......

Have fun calibrating.

 

[bushisland]

I will, but more than that is I have learned a lot from this. Thanks again for your time and patience.
Best Regards,
bushisland