"Core Temp" - Official Developer Thread.

[Bill Brasky]

For some reason I've always preferred the layout of coretemp, and I like that the tray icon is configurable.

 

[The Coolest]

Version 0.99.8.0 - 3rd October, 2010

- New: Completely new settings dialog, much clearer and easier to use.
- New: Added Hibernation to Overheat protection options.

- Fix: Mobile Intel Core2 Quad processor recognition issues.
- Fix: Pentium Dual-Core E6000 series model detection was incorrect.

- Update: TjMax adjusted to 100c (Intel's specs) for 400 and 500 series Atom processors.

 

[Idontcare]

Hi, I notice you are making TjMax updates in CoreTemp to properly reflect Intel's specs.

Any chance in updating CoreTemp to use the correct TjMax for the G0 Q6600?

Currently CoreTemp uses 100°C as the TjMax but the correct TjMax for the G0 is 90°C per Intel's specs.

Spec was published by Intel in 2006 at their fall IDF. See page 15 of http://lenry.atw.hu/tjmax.pdf

 

[ilkhan]

My i5-540M doesn't display the VID.

 

[bryanW1995]

what about thuban? tjmax is listed as 70c on my 1055t but I ran it as high as 88c (before using the 15c offset that was recommended) overnight with zero thermal throttling, performance degradation or problems of any kind. I'm thinkin' that might be a bit off.

Is there a way to manually enter tjmax?

 

[iCyborg]

My i5-540M doesn't display the VID.

I believe no program displays VID for any Core iX

 

[The Coolest]

Hi, I notice you are making TjMax updates in CoreTemp to properly reflect Intel's specs.

Any chance in updating CoreTemp to use the correct TjMax for the G0 Q6600?

Currently CoreTemp uses 100°C as the TjMax but the correct TjMax for the G0 is 90°C per Intel's specs.

Spec was published by Intel in 2006 at their fall IDF. See page 15 of http://lenry.atw.hu/tjmax.pdf

This was discussed in length a couple of years ago, mostly on ocforums.com and XS, when Intel has just released its specs.
They are clearly off by a big margin. Their 45nm Core 2 specs were pretty accurate, but just didn't make sense for the 65nm chips, where most users were reporting idle temps way below ambient.

My i5-540M doesn't display the VID.

As iCyborg said, the new generation of Core CPUs do not report VID back to applications, so it can't be read. VID and VCore are not the same, Core Temp's FAQ explains it in section #14 & #15.

what about thuban? tjmax is listed as 70c on my 1055t but I ran it as high as 88c (before using the 15c offset that was recommended) overnight with zero thermal throttling, performance degradation or problems of any kind. I'm thinkin' that might be a bit off.

Is there a way to manually enter tjmax?

The TjMax in the K10\K10.5 is encoded into the processor's registers, it should react similar to how Intel CPU's react when reaching TjMax, but I don't know at what exact point that happens. Also, adjusting TjMax is currently not possible, it also doesn't do anything for you except see a higher digit.

 

[Idontcare]

This was discussed in length a couple of years ago, mostly on ocforums.com and XS, when Intel has just released its specs.
They are clearly off by a big margin. Their 45nm Core 2 specs were pretty accurate, but just didn't make sense for the 65nm chips, where most users were reporting idle temps way below ambient.

Its not clear whether you are saying you believe the Intel spec itself is wrong (i.e. you claim TjMax of 90°C itself is wrong) or whether you are saying the DTS slope error is large and thus results in the reported temps at idle being "off by a big margin".

(slide 8 of the Intel IDF document)

In either event, the spec value really is 90C. It doesn't make much sense to knowingly use the incorrect TjMax just to make the idle temps "seem" to make sense when the DTS error is still present regardless.

Unless you know the DTS slope error, the temps coretemp produces are going to be wrong anyways, but at least they become less wrong as the loaded temps approach TjMax.

But that isn't true if you use the wrong TjMax. In that case your reported temps continue to be wrong even as the CPU is loaded, and at TjMax they will currently be wrong by 10°C.

Personally I monitor my CPU temps because I am interested in knowing how hot things get when my system is loaded, that is when thermal instability can be an issue and that is where I optimize stability versus Vcc versus fanspeed. I don't want my reported temps to be intentionally wrong on top of already being slightly in error from an unknown value of DTS slope error.

Personally I'd like to see Coretemp use and report the correct TjMax value for my G0 by default and if some people have DTS slope error that makes them uncomfortable with their reported idle temps then they should elect to change the offset by an amount that meets their individualized estimation of the DTS slope error rather than the program just generically assuming everyone's Q6600 G0 has the same 10°C slope error which is "over-corrected" by way of changing the TjMax value to a non-spec value.

 

[The Coolest]

You bring a valid point. But there is a but:
Most people who use temperature monitor applications will not bother reading up on them or the sensors, or look into how they work in reality. They won't bother with sensor calibration, and they won't appreciate "inaccurate" readings that make no common sense. Most people want stuff that you click and see the result right away without much messing around. Even though I'm a developer, and consider myself an enthusiast, I still notice that for most hardware and applications that I use, I prefer to install and go, without spending hours configuring it to work right from the get go.

The slope error of the sensor is problematic, since it's not a constant, there are no numeric ranges to allow Core Temp to accurately or even approximately calculate the correct temp distance from the TjMax.
In my personal opinion, the best way to monitor temperature would be to look at the raw TjMax figure. And as long as it's 20C+ away from 0 you're perfectly fine. Most people would find this unnatural and confusing, and they would prefer to see absolute values.

There are other apps, such as RealTemp, which allows sensor calibration, for example. I'm not sure what TjMax values are used in it exactly, but I'm pretty sure that they are also not the ones posted in that Intel document.

 

[Lazlo Panaflex]

Originally Posted by The Coolest

The TjMax in the K10\K10.5 is encoded into the processor's registers, it should react similar to how Intel CPU's react when reaching TjMax, but I don't know at what exact point that happens. Also, adjusting TjMax is currently not possible, it also doesn't do anything for you except see a higher digit.

Q - so if my Thuban @ 3.1 is running @ 46c, full load (Prime 95, small FFT's), am I really 16c away from throttling? (AMD says the Thubans should go no higher than 62c.)

 

[The Coolest]

That's exactly how I understood it from their tech doc. Of course if you've manually added an offset, the offset should be ignored/subtracted from the temperature Core Temp displays.

 

[Lazlo Panaflex]

Nope, no offset. I just go with what the program tells me

Edit: thanks for the new version, btw

 

[The Coolest]

Hey guys...
I know it took forever, but Core Temp version 1.0 (RC1) is here at last!
There were many delays due to life, college projects and other things that just got in the way.
I hope that the improvements and enhancements which v1.0 brings will make up for it.
So without further ado here is the change log for the new version.

Version 1.0 RC1 - 18th August, 2011

*** Core Temp Monitor is now available, an Android app to monitor your machines from anywhere in the world!
More information is available here

- New: Multilingual interface. Core Temp now natively supports adding non-English languages.
- New: Plugin system - 3rd party developers can now create plugins and extensions for Core Temp, both native code and .Net is supported, please see this for more information.
- New: Intel Sandy Bridge and AMD Fusion (Brazos, Llano) support.
- New: Max TDP detection on supporting processors.
- New: Power consumption on Intel's Nehalem and newer processors.
- New: Added support for VIA processors. (C3 - detection only, Temperature by Acpi Thermalzones)
- New: Added support for Intel Pentium and AMD K5 and newer processors. (detection only, Temperature by Acpi Thermalzones)

- Fix: All of the user reported bugs and many more unreported problems.

- Update: Optimized the startup code, Core Temp should now launch 2 to 4 times faster.
- Update: Optimized many other aspects of the code, Core Temp should now consume much less CPU time than before.
- Update: Restored support for older OS: Windows XP is now fully supported, Windows 2000 requires security updates KB935839, KB835732 and GDI+. In some cases hotfix KB816542 may be needed as well.

 

[E6700]

all that trouble just to remote core temp? no thankyou

 

[The Coolest]

What is the trouble?
All you need is to install the plug-in and forward a port in your router.

 

[Dadofamunky]

What about modern Sandy Bridge support? Congratulations, by the way

 

[grimpr]

Fantastic release, thanks TC!

 

[The Coolest]

@ Dadofamunky:
Sandy Bridge is now fully supported, with TDP and current power usage readings.

@ grimpr
Thanks

 

[Capt Caveman]

Awesome, thanks!

 

[Thor86]

Excellent tool, but where is the stand-alone executable for v1 RC1? Don't feel like installing.

 

[MadScientist]

Excellent tool, but where is the stand-alone executable for v1 RC1? Don't feel like installing.

Click on his link. Under: Download: Core Temp 1.0 RC1 click on Extras.

 

[bryanW1995]

First of all Core Temp has support for AMD processors, RealTemp doesn't.
Besides that Core Temp also has a few addons that RealTemp doesn't, you can check them out here:
http://www.alcpu.com/CoreTemp/addons.html

Other than that it has a few small features like Window 7 taskbar support and several other small additions that RealTemp may not have implemented.
Of course there are several things that RealTemp has which Core Temp doesn't.

I think that both programs are mostly an acquired taste, try it and see if you like it better than RealTemp or not, you can always delete it if you don't want to keep using it.

Core temp isn't bad, but I only use it on my Thuban b/c realtemp, as mentioned, doesn't support amd cpus. Maybe if BD ends up being a hit Kevin will take the time to add in the support, but I wouldn't hold my breath at this point.

edit: I also use it on my laptop, plus this new program looks pretty good. Maybe I'll go back to coretemp on all my rigs now.

You bring a valid point. But there is a but:
Most people who use temperature monitor applications will not bother reading up on them or the sensors, or look into how they work in reality. They won't bother with sensor calibration, and they won't appreciate "inaccurate" readings that make no common sense. Most people want stuff that you click and see the result right away without much messing around. Even though I'm a developer, and consider myself an enthusiast, I still notice that for most hardware and applications that I use, I prefer to install and go, without spending hours configuring it to work right from the get go.

The slope error of the sensor is problematic, since it's not a constant, there are no numeric ranges to allow Core Temp to accurately or even approximately calculate the correct temp distance from the TjMax.
In my personal opinion, the best way to monitor temperature would be to look at the raw TjMax figure. And as long as it's 20C+ away from 0 you're perfectly fine. Most people would find this unnatural and confusing, and they would prefer to see absolute values.

There are other apps, such as RealTemp, which allows sensor calibration, for example. I'm not sure what TjMax values are used in it exactly, but I'm pretty sure that they are also not the ones posted in that Intel document.

I've said this for years. Of course, I still don't do it, either... I'm making a new years eve vow to start it in 2012, however!

btw, how do I set up coretemp to register delta to tjmax? I just fiddled with it for a few minutes without success. Of course, Ive been running over the alleged tjmax of 70 for quite some time. I just looked at my high temp (88c), even with the possible 15c offset that I've seen/heard about I've still gotten 3c over tjmax, but I'm still 100% stable (as in running seti 24/7/365 on this rig + gpu with zero bad WU's).

 

[MadScientist]

Buggy?
Under load, it will for a second flash a yellow 98C? temp and record this as the max temp for that core. Does it for all cores.
The watts jump all over the place, from 11.2W to 1200W.
Compared to V0.99.8 temps are ~the same.

[/IMG]

[/IMG]

 

[The Coolest]

...
I've said this for years. Of course, I still don't do it, either... I'm making a new years eve vow to start it in 2012, however!
...

Something to consider: Since the original Nehalem, the TjMax is pre-programmed into the CPU. This makes the guessing games we had with Core 2 irrelevant. It may not be 100% accurate, but it's probably the closest you can get without going to extreme measures to obtain an accurate reading.
The statement you quoted is less valid today than it was a few years ago.

Buggy?
Under load, it will for a second flash a yellow 98C? temp and record this as the max temp for that core. Does it for all cores.
The watts jump all over the place, from 11.2W to 1200W.
Compared to V0.99.8 temps are ~the same.

IMG]http://i164.photobucket.com/albums/u30/bomart4748/Coretemp2.png[/IMG]

IMG]http://i164.photobucket.com/albums/u30/bomart4748/Coretemp.png[/IMG][/IMG]

That is indeed a new bug. I'm working on getting it fixed.

 

[Stuka87]

Woohoo, thanks for adding SB support