Real Temp - New temp program for Intel Core processors

[Talonman]

I agree. I guess for the better, or for the worse is where the pack splits.

[Demo]

But its good to have them all lined up when loaded, though i need more than a 3 for max calibration.. 1 of my cores is 9c lower when loaded.
It does not bother me much since im always looking at the TjMax, if i see it goes 23 then thats danger for me XD

[Talonman]

The thing that kicks my butt is the core that reads too low at idle, will for sure be the same core that will read the lowest under load too. (With no calibration)

I know the sensors get more accurate under load, but the fact it happens at idle and loaded, kind of supports that you may have a sensor that just likes to report low by it's nature.

A small calibration fixes this right up for me. I know it makes my idle temps more accurate. The only question I have is could it also be making my loaded temps more accurate too?

I'm thinking yes. Manual compensation for an anemick sensor is a positive thing.

There might be allot of people thinking they had uneven pressure on their block, went for a remount, and the same core still reported lower than the rest. I would chalk that up to 'Sensor reads low' rather than 'poor mount' myself...

If you have good cooling, and calibrate each core at idle to read around 6.5°C above your ambient, your loaded temps will make your mount look like a pro did it by how even they are.

[Demo]

Well the paste behind the IHS and its contact also matters, a friend of mine removed the IHS of his Q6600 and the CPU ran 9c cooler ( all 4 cores ) and his heatsink is a lot warmer to the touch now even at lower temps.. this shows that its getting almost perfect surface contact.

[Demo]

uncle, can you add a PWM temp monitor in real temp ?

[unclewebb]

Yeah thats the problem, Load temps get affected by the calibration.

That isn't because I screwed up or anything like that. It is by design. When a sensor is out by 8C or 9C at idle, like many of them are, that error doesn't just disappear the moment you put a load on your CPU. The errors in these on chip sensors doesn't completely disappear until you hit approximately 60C. Users that are water cooled and don't ever get up to this core temperature are still benefiting from the calibration they did at idle.

If I come over to your house with my IR thermometer and pull off your heatsink and risk damaging your CPU, I can probably set-up RealTemp so that your reported temperatures are within +/- 1C of the actual temperature. By following my calibration guide lines instead, you'll probably end up at about +/- 2C from idle to 60C and +/- 1C from 60C to TjMax without having to risk anything. That's far better than any of the competition can give you.

Maybe I should rename that feature to Sensor Calibration so there is less confusion. My calibration test involves bringing your CPU down to it's lowest temperature at low MHz and low core voltage but that's just a convenient reference point. This simple calibration done at idle will improve the accuracy of your reported core temperatures from this point all the way up to 60C. From about 60C to 95C these sensors do an excellent job of tracking changes in core temperature so no further calibration adjustment is needed.

Talonman: If you liked previous versions of RealTemp then you are definitely going to be happy with the beta release coming out later today. Only a few very minor things left to do.

uncle, can you add a PWM temp monitor in real temp ?

No. RealTemp reads the Intel on chip sensors and that's pretty much it.

I dont really use the load calibration, its like fooling my self that all my cores are running at equal temps. If they really are running equally the delta to TjMax should be equal also.

The delta to TjMax would be equal for any two cores if the sensors were working properly but they're not. It is common to find temperature differences between the two pairs of cores within a Quad core processor at full load due to heat transfer issues but two cores within the same dual core should be running at the same temperature when at full Prime / Orthos small FFT load. If they are not reporting the same temperature then it is very likely to be a sensor issue.

[blunden]

Just assuming the TjMax is 95 C instead of 100 C seems a little risky and a little like wishful thinking to me. Believe me, I would love to have 5 C lower temps. I know about your tests (I read at least 20 of the pages in this thread) and I can't think of any better way to test it myself. Though, how can we be (close to) certain this is correct? How was the 100 C value calculated? Is has to be more than a guess, right?

Is the general consensus that a TjMax at 95 C is the most accurate value for the Q6600 G0 at the moment, or is it only so in this thread?

[Talonman]

I would say the general consensus of this thread is we don't believe idle temp's that are reported too close to ambient to be believed, and view Sensor Calibration is a positive thing.

A sensor that dosen't give accurate info until 60°C or higher, is not a watercooled PC's friend. Better to have my manually calibrated sensors mabey reading a tad high, or even dead balls accurate, rather than knowing it's reading unbelievably low and choosing to believe that as truth.

You have to pick a camp.

INTEL insured we can't have both.

Talonman: If you liked previous versions of RealTemp then you are definitely going to be happy with the beta release coming out later today. Only a few very minor things left to do.

Looking foward to it!

[rge]

How was the 100 C value calculated? Is has to be more than a guess, right?

100C was not calculated. Some 65nm mobile cpus have a documented tj of 100, so by guessing 100 for Q6600 even though it is a desktop quad and not a mobile cpu, it was hoped to be in the ballpark.

For example. 45nm mobile cpus have a documented Tj of 105 in intel spec sheets. Even though intel has stated it is not correct to use mobile cpu tj for desktops and even though mobile cpu's have no IHS and different cooling solutions and thus reasonable to have a different Tj, prior to unclewebbs program, guessing 105 for tjmax of 45nm desktop cpus and hoping it was in the ballpark was all one had to go on. Though speedfan dissented and just kept 100 tjmax for 45nm desktops, possibly (though I am guessing) because 105 tjmax gave nonsensical temps.

[blunden]

100C was not calculated. Some 65nm mobile cpus have a documented tj of 100, so by guessing 100 for Q6600 even though it is a desktop quad and not a mobile cpu, it was hoped to be in the ballpark.

For example. 45nm mobile cpus have a documented Tj of 105 in intel spec sheets. Even though intel has stated it is not correct to use mobile cpu tj for desktops and even though mobile cpu's have no IHS and different cooling solutions and thus reasonable to have a different Tj, prior to unclewebbs program, guessing 105 for tjmax of 45nm desktop cpus and hoping it was in the ballpark was all one had to go on. Though speedfan dissented and just kept 100 tjmax for 45nm desktops, possibly (though I am guessing) because 105 tjmax gave nonsensical temps.

Good to know. Did anyone try Unclewebb's measuring method on a mobile CPU yet with known TjMax? Didn't see any in all the pages I read but I guess I might have missed it.

[Demo]

That isn't because I screwed up or anything like that. It is by design. When a sensor is out by 8C or 9C at idle, like many of them are, that error doesn't just disappear the moment you put a load on your CPU. The errors in these on chip sensors doesn't completely disappear until you hit approximately 60C. Users that are water cooled and don't ever get up to this core temperature are still benefiting from the calibration they did at idle.

If I come over to your house with my IR thermometer and pull off your heatsink and risk damaging your CPU, I can probably set-up RealTemp so that your reported temperatures are within +/- 1C of the actual temperature. By following my calibration guide lines instead, you'll probably end up at about +/- 2C from idle to 60C and +/- 1C from 60C to TjMax without having to risk anything. That's far better than any of the competition can give you.

Maybe I should rename that feature to Sensor Calibration so there is less confusion. My calibration test involves bringing your CPU down to it's lowest temperature at low MHz and low core voltage but that's just a convenient reference point. This simple calibration done at idle will improve the accuracy of your reported core temperatures from this point all the way up to 60C. From about 60C to 95C these sensors do an excellent job of tracking changes in core temperature so no further calibration adjustment is needed.

Talonman: If you liked previous versions of RealTemp then you are definitely going to be happy with the beta release coming out later today. Only a few very minor things left to do.



No. RealTemp reads the Intel on chip sensors and that's pretty much it.



The delta to TjMax would be equal for any two cores if the sensors were working properly but they're not. It is common to find temperature differences between the two pairs of cores within a Quad core processor at full load due to heat transfer issues but two cores within the same dual core should be running at the same temperature when at full Prime / Orthos small FFT load. If they are not reporting the same temperature then it is very likely to be a sensor issue.

Thanks for the enlightenment,
My distance to TjMax usually looks like this
35-34-40-38

Seem like either a sensor or a mounting problem.

[Richard Dower]

Any new betas out for download?

[Demo]

Lower my fan speed to increase the temps
im getting 29-29-33-32 on delta to TjMax
Thats 66-66-62-62 w/ calibration enabled, so i guess the accurate point is at delta 33 and lower..

[unclewebb]

Somewhere in this long thread I showed that even with properly functioning sensors, a Quad core at full load will show two cores running at one temperature and two cores running at a different temperature. My conclusion was that this is most likely caused by heat transfer issues. When I turned my fan completely off while running Prime the difference between the two sets of cores in my Q6600 increased to about 10C. The normal amount for me was a difference of about 5C while running Prime.

At idle, at the same temperatures as above, the reported temperatures from both sets of cores was much closer to equal which to me showed that it's not the sensors causing this issue but heat transfer, either IHS to heatsink or from the cores to the IHS. When I totally loosened up my heatsink and this difference was still there, I concluded that it must be the cores to the IHS where the problem is.

Any new betas out for download?

One more button to hook up in the Settings window. This new beta version has been thoroughly gone over but it seems to be working as planned. We're down to the final hour before launch.

blunden: Did anyone try Unclewebb's measuring method on a mobile CPU yet with known TjMax?

I haven't done this test yet or heard of anyone else doing it. I'm not 100% convinced that mobile chips have a "known TjMax" regardless of what Intel documentation says. Too much information on Intel's own forums and website has mysteriously disappeared that I don't believe anything Intel has to say when it comes to the subject of TjMax. I did test my Q6600 - G0 and it sure looked like TjMax=95C was correct. It was a brand new processor so I didn't do a thorough test but hopefully in the near future I can have a second look at it without a heatsink getting in the way. I remember posting two pictures of that test somewhere in this thread but not sure where. Too busy programming lately!

[blunden]

I did test my Q6600 - G0 and it sure looked like TjMax=95C was correct. It was a brand new processor so I didn't do a thorough test but hopefully in the near future I can have a second look at it without a heatsink getting in the way. I remember posting two pictures of that test somewhere in this thread but not sure where. Too busy programming lately!

Yeah, went back and read 5-6 extra pages. I think I saw you mentioning it at page 43 or so. Did you mean further back? If so, I will search for it when I have the time.

[jas420221]

Just a quick FYI... I wouldnt reference things in a 'page' fashion. For example, there are only 39 pages in this thread to me (b/c I made changes in my usercp). Post # would be a lot better for those that (may) have changed their settings to view more posts /page.

[unclewebb]

Beta days are here again.

http://www.fileden.com/files/2008/3/...alTempBeta.zip

Unzip and copy the beta files into your RealTemp directory and install the RTFont in the Font directory. Right click on the RTFont file in Vista or open the Control Panel -> Font directory in XP and drag it in there to install it.

There have been some significant changes but I'm hoping that most of the bugs have already been worked out.

In the new Settings window the TjMax and Idle settings can be adjusted on the fly. Just tab to the next box and the change should take effect immediately. Same with the tray fonts. The box on the far right is for when you're using the RTFont. If the RTFont is not installed the results might be 4 blank icons in the tray or 4 ugly looking ones. I made a couple of slight changes to this font so if you were using it previously you might want to uninstall your old version and install this one instead. The changes are very minor.

I haven't had a chance to work on the MHz calculation yet. If CPU MHz isn't checked and you are not using SetFSB then it will work fine as is. Select this option if you are using SetFSB. If it doesn't work 100% correctly you can PM me or wait until I have a chance to fix it. It already works properly on my motherboard when using SetFSB. It's the next item I plan to work on some more for better compatibility.

I'm hoping the new user interface is fairly intuitive and user friendly. What do you think? Any problems? Will this new version encourage some more CoreTemp users to change sides?

[camouflage]

Nice work uncle - all is running fine so far, but I miss the button on the top right corner to switch from MHZ to VID and so on......

[unclewebb]

Nice work uncle - all is running fine so far, but I miss the button on the top right corner to switch from MHZ to VID and so on......

Thanks. Unfortunately, progress always has its casualties. You can just as easily click on the Settings button and see all that information. I will return the little button to the main page if I get enough requests for it.

[camouflage]

Thanks. Unfortunately, progress always has its casualties. You can just as easily click on the Settings button and see all that information. I will return the little button to the main page if I get enough requests for it.

No problem - missing the button won't drive me into suicide......

[blunden]

Just a quick FYI... I wouldnt reference things in a 'page' fashion. For example, there are only 39 pages in this thread to me (b/c I made changes in my usercp). Post # would be a lot better for those that (may) have changed their settings to view more posts /page.

I only referenced it as being on a specific page because I didn't remember exactly. Found it again and it was post #1074, but I now saw that it can't be the one he was talking about. It's too short and no pictures. Probably just a reference.

http://www.xtremesystems.org/forums/...postcount=1074

[XtremeTiramisu]

Beta tested the 2.56 real temp and it's working great.
I can easily activate all four cores from the setting tab and there are four font options availiable.
I actually like the preset font1 the best and the rest are..well..it's matter of personal preferences.
Now, wel'll wait for the font color options to be availiable from the setting tab to make it even more sweeter.

Thanks Unclewebb

[FullSky]

Testing Beta 2.56

One oddity I found was on the Settings dialog box, System Tray Settings group. Depending on the order in which you click the Display Core check boxes also determines the order in which the core temperatures appear in the system tray.

This changing order of the core temp display in the system tray can be confusing. I think it would be better if the order is always core 0, 1, 2, and 3.

[fgw]

thanks for your effort uncleweb!

here is my input:

settings gui:system tray settings
agree to fullsky: you should change the order how temps of individual cores are displayed in the tray to a fixed order. currently the order depends on the order of selecting the cores in the setup gui

what is the "font options" button for?

tray
moving the mouse over the temps in the tray shows "core 0: 35C" and "core 1: 35C" if i point to the left tempfigures. pointing to the right tempfigures the display looks like "core 1". running on a dualcore, no quad here. unchecking one core in the settings gui displays the correct display of "core 0: 35C" and "core 1: 35C" when pointing to it in the tray.

[unclewebb]

I agree that if you go into the Settings window and click on cores randomly that the order in the System Tray area can be confusing. The only way around this is each time a user selects or removes a core from the System Tray, all the tray icons would have to be deleted and then all of the Tray icons would have to be re-built and added back to the system tray in the proper order, one by one. I'll try programming it this way to see how it looks. Unfortunately there doesn't seem to be a way to insert or remove a tray icon between two existing tray icons.

fgw: I presently have RealTemp set to show the temps of all cores when you mouse over the left most tray icon and then to show the core number for the remaining tray icons. I'll set it to show the temperatures of all cores over each tray icon and see if I get more or The Font option should change the size and style of the tray font used.

XtremeTiramisu: Now, we'll wait for the font color options to be availiable from the setting tab to make it even more sweeter.

If you click on the buttons that say Core 0, Core 1, etc. in the Settings window you should be able to adjust the font color of your tray icons. Is this not working for you or .....?