Re: Suggestions wanted

[Paul Goyette <paul%whooppee.com@localhost>]

On Sun, 5 Oct 2008, Paul Goyette wrote:

> I would expect that the CPU does "know" its Tcc value, but it's not clear 
> that there's any way to ask it, other than maybe another command over the 
> PECI interface.  The ADT7490 doesn't provide any generic PECI access, and (at 
> least so far) there's nothing else in the NetBSD repertoire that provides 
> such access, either.
>
> I've got an inquiry in to ON Semi's support center to see if there's a hidden 
> access method via the ADT7490, and I'll be scouring the Intel spec sheets to 
> see if there's anything available that could be accessed as part of our CPU 
> identification code.

I got confirmation from ON Semi today.  Not only is there no way to 
determine the CPU's Tcc from the ADT7490, but as far as they know 
there's no way for _any_ PECI master to get that info from the CPU. 
(This is confirmed by various blogs and threads I've run across.)

Apparently, Intel in their infinite wisdom provides BIOS vendors a 
detailed list of CPUs, indexed by type and stepping, and the related 
value for Tcc.  Then it's up to the BIOS vendors to adjust temperature 
readings based on this reference value.  Damn, I hate proprietary 
solutions to generic problems!  :)

So, unless there's suddenly a change on Intel's part, and/or some 
significant cooperation between BIOS vendors to provide a documented 
access method, I think we're going to be stuck with adding a "relative 
temp" sensor type to sysmon_envsys().  I'll be adding this to the 
enhancement/modification list I'm preparing to work on in the post-5.0 
timeframe.


----------------------------------------------------------------------
|   Paul Goyette   | PGP DSS Key fingerprint: |  E-mail addresses:   |
| Customer Service | FA29 0E3B 35AF E8AE 6651 |  paul%whooppee.com@localhost   |
| Network Engineer | 0786 F758 55DE 53BA 7731 | pgoyette%juniper.net@localhost |
----------------------------------------------------------------------