> Can we not have so many magic numbers in the code, or at least comments
> explaining what the bit values mean?
>   
those magic numbers are pentium III cpuids which support split voltage 
and therefore should work with SMI,
maybe there are more but the else part is only triggered if the BIOS 
couldn't set some correct values.
Sadly I couldn't find a better solution, there is no feature-flag for 
the first generation SMI.
It would be also possible to remove the complete else-switch thing and 
estimate: if the BIOS don't set SMI,  we also don't use it. So the magic 
numbers are gone and the system don't do bogus things on PIIX4 systems 
with non-pentium III processors attached. But at least all other OS I 
checked have a fallback solution. While the Linux one is maybe the most 
configurable, because the user can setup the default (no hardcoded values)

Best Regards
Stephan