Subject: speedstep and the 133mhz fsb chips
To: None <port-i386@netbsd.org>
From: Tom Spindler <dogcow@babymeat.com>
List: port-i386
Date: 12/28/2005 14:44:43
My shiny new Pentium-M 1.86Ghz laptop isn't supported by the current
incarnation of NetBSD's SpeedStep support; in poking around, it appears
that two things are happening:

1. http://support.intel.com/design/mobile/datashts/305262.htm
   points out that BSEL[1:0] indicates the FSB multiplier; currently, the
   speedstep code assumes that the frequency will always increase in units
   of 100Mhz. This isn't the case for some of the newer chips - they use
   133Mhz instead (which explains why my 1.86Ghz chip shows up as 1.4Ghz.)

2. http://www.intel.com/support/processors/mobile/pm/sb/cs-007981.htm
   lists the voltages for the 90nm-process chips; somewhat vexingly, while
   it gives a fixed voltage for the 'battery optimized mode', this
   changes for the lowest-powered chips - and the 'maximum performance'
   is a big ol' range of voltages (save for the 778, which has a lower
   maximum performance voltage than any of the rest.)

Since I don't know doodly about how the chipsets are supposed to work -
I'm inferring that you need to figure out BSEL from the chipset, rather
than the CPU - would one of you kind folks be able to make stuff go?
It also adds the complication that you can have different processors
with the same final speed and possibly requiring separate voltages.