CPU Temperature incorrectly reported for Atom N270 (too low)

[GoogleCodeExporter]

CPU Temperature is reported as very low, sometimes (at PC startup) even under 0 
°C. Based on some experimentation, comparing with HWMonitor Pro, I've found 
TjMax should be set to 105 for this CPU (instead on 90).
Same problem on RealTemp.

What version of the product are you using? On what operating system?
0.1.35 on Windows 7 Starter edition (32 bit)

Please attach a Report created with "File / Save Report...".
Attached report, with CPU Family and ID.

Original issue reported on code.google.com by andreabe...@gmail.com on 6 Oct 2010 at 9:57

Attachments:

• OpenHardwareMonitor.Report.txt

[GoogleCodeExporter]

TjMax for this CPU is in fact 90, as can be verified at 
http://ark.intel.com/Product.aspx?id=36331

The problem with the temperature having inaccurate values can have several 
causes. We'll dig into this (as well as looking if we can add support for the 
chips used on your board), can you in the meantime upgrade to the latest 
version (0.2.1, available from the downloads section) and upload a report from 
that version as well?

Thanks for your report though, gives us some more insight into these Atoms!

Original comment by paulwere...@gmail.com on 6 Oct 2010 at 10:39

• Changed state: NeedsInfo

[GoogleCodeExporter]

You can't really determine the correct TjMax from idle temperatures, because 
there is also the slope error. The only way would be to run the CPU (carefully) 
without cooler, and measure the temperature on the top of the heat-spreader 
while operating the CPU near TjMax (distance from TjMax = 0). This should at 
least give a lower limit for the TjMax, because the core can't really be cooler 
than the heat-spreader. But it's not even clear if TjMax is the same on all 
CPUs of the same type. So this would be valid only for the CPU its measured on.

I have reports from two other systems using an N270 (both model 0x1C, stepping 
0x2):

+-+ Intel Atom N270 (/intelcpu/0)
|   +- Clock[0] : 133.3705 : 133.3705 : 133.3705 : Bus Speed
|   +- Clock[1] : 800.2227 : 800.2227 : 1600.445 : CPU Core
|   +- Temperature[0] : 43 : 42 : 50 : CPU Core
|   +- Load[1] : 3.125 : 0.78125 : 68.75 : CPU Core

+-+ Intel Atom N270 (/intelcpu/0)
|   +- Clock[0] : 133.0401 : 133.0401 : 133.0401 : Bus Speed
|   +- Clock[1] : 798.2407 : 798.2407 : 1596.481 : CPU Core
|   +- Temperature[0] : 40 : 38 : 48 : CPU Core
|   +- Load[1] : 16.5896 : 1.543462 : 100 : CPU Core

Here the temperatures look not too wrong. But again, idle temperatures don't 
say much about the TjMax.

Even with a TjMax of 105°C, your CPU core temperature would be 11°C after 
starting Windows. That's something I can't really believe to be correct, unless 
you use sub-ambient cooling.

Without more evidence that the TjMax of 90°C is really wrong, we can't do much.

Original comment by moel.mich on 6 Oct 2010 at 11:53

[GoogleCodeExporter]

OK guys, it seems i'm falling in incorrect values reading, as described in: 
http://www.tomshardware.com/news/intel-dts-specs,6517.html.

I've found many posts on internet about relations between Intel official 
Tjunction value and TjMax and it seems not so clear. Someone says TjMax is 
based on silicon measurement during production process, so, it can be really 
different from CPU to CPU.

Anyway, having a look at: 
http://macles.blogspot.com/2008/08/reading-internal-thermal-diode-of-intel.html 
seems temperature on Atoms should not be trusted when under 50°C (like you 
already noted).

Maybe, as a relatively easy way to alleviate offsets between temperatures, it's 
possibile to add to OHM an "Advanced Calibration Settings" as in RealTemp (as 
seen here: http://img188.imageshack.us/img188/5448/picxu.png).

I'm attaching OHM report from latest version.

Original comment by andreabe...@gmail.com on 6 Oct 2010 at 12:56

Attachments:

• OpenHardwareMonitor.Report.txt

[GoogleCodeExporter]

There are two calibration parameters for Intel CPU core temperature sensors 
which can be edited by the user: TjMax and TSlope. The temperature shown in the 
user interface is then calculated using the forumla

  Temperature = TjMax - TSlope * Value

where Value is the raw value read from the sensor. You can edit the parameters 
by double clicking the sensor or right clicking and selecting "Parameters...". 

Original comment by moel.mich on 6 Oct 2010 at 1:44

[GoogleCodeExporter]

On my netbook the TjMax must be 105 to get a value that corresponds with 
temperatures of other monitoring tools.

PS: Looks like the edit field for the TSlope parameter accepts int values only, 
is this correct, the parameter seems to be a float in the code?

Original comment by jubel...@googlemail.com on 14 Nov 2010 at 11:23