Sensor reading on a Dell 860

[dutchadept]

Hi Martin,
First off my thanks for your great app. I've been using Speedfan and Hmonitor but both programs just don't seem to cut it anymore.
I just migrated my homeserver platform to a Dell 860 server and currently trying to get all the sensor readings into my home automation (on win 7 x64).
The board is based on the intel 3000 chipset (with ICH7) and uses a LM96000 IO.
Some reading in HWiNFO are correct (FAN Rpm, backup battery voltage) but my processor temp needs an offset (-51°C is an overclockers dream but slighly unrealistic). I couldn't find any setting to correct this.
Also planar (board) temp is rock solid 25°C, and eventhough this system runs exceedingly cool, it's also impossible to have no variation in board temp.
Any chance to either fix up the readings or add an offset variable to the program?

Thanks,

Alex

(debug attached)

 

[Martin]

Hi Alex,

you haven't attached a report file, but according to the debug file I assume you're talking about the IPMI sensor (the LM96000 doesn't seem to provide any readings).
I have checked the IPMI configuration and it really seems that the system returns a negative value. The raw value read via IPMI gives x=76. Then the final equation (according to how the IPMI system is configured) is: y = Mx + B (where M=1 and B=-127), so that results in 76 - 128 = -52.
I've seen such behaviour on some systems and it seems that the CPU temperature returned by the IPMI is the distance to a marginal value (like Tj,max in DTS).
I'm not sure if it's possible get this marginal value via IPMI or how to convert it to real temperature.
Please try one test in order to verify if I'm correct - put load on the CPU and watch how this value behaves.
You might also try some other IPMI viewer tool to verify the value.

Regards,
Martin

 

[dutchadept]

Hi Martin,
Yes, I had the same notion regarding the CPU temperature reading.
I ran some stress tests and the IPMI CPU temp followed the CPU load curve nicely. So all I need to do is use my digital thermometer to measure the actual temp on the die to get an offset number. I would appreciate it if you could include an temp offset variable in the .ini file of a future release to correct the reading.
As for the National Semiconductor, I wonder if it really doesn't produce any readings. If I remember correctly, this unit could have it's base address changed (to be able to use multiple units), so it might be that Dell decided to use a higher address (even though only one unit is implemented).

Anyway, thanks for your quick reply!

Alex

 

[Martin]

The LM96000 was found on SMBus by HWiNFO32, however the reading registers do not report any value.
The IPMI CPU temperature might be an offset from the same Tj,max as the CPU DTS is (in your case it should be 85 C probably). But this needs to be proved.
Also depends on implementation - in case this machine reports the same CPU DTS temperature via IPMI too, then you could simply ignore the IPMI reading and use the DTS reading provided by HWiNFO32 instead. You can also adjust the DTS Tj,max in case you think it's not correct (just right-click on the DTS reading).