[ltp] T40: installation report

[Neil Weisenfeld]

Hi Fabrice,

Great information on your web page!

I'm interested in your memory errors.  I'm having the same thing.

I also have an odd problem with my T40p since the beginning and I've
tried a number of different things.  I think that the culprit may be
these memory errors.  It will be interesting to see if it's truly a
problem or a memtest bug/incompatibility.

The issue I'm having is random crashes of almost every application on my
machine.  From the X server to the kernel itself to various panel
applets, etc.  About 5 times a day something just folds up and
disappears for no reason.  I pulled the 512MB chip and it will be
interesting to see how tomorrow goes.

Re: the built-in wireless (not that you asked me, but it was on your web
page ;-).  I don't think that the open-source project is going to really
get anywhere anytime soon, though I'd love to see it happen.  From what
I've read, the Atheros chipset is basically the WinModem of wireless --
most of the work is done in the drivers, so that implementing them is
far from trivial.  They also, apparently, already sent their lawyers
after the guy heading up the open-source driver effort, although he's
outside of the US and seems not to care very much (he also seems to know
what he's doing).  Someone (Andy Ts'o?) talks on his web page about
pulling the mini-PCI card and replacing it w/ a Prism-II based model and
that may end up being the easiest path for now.


Best regards and thanks for writing up all that you did.



Neil





On Thu, 2003-06-12 at 08:42, Fabrice Bellet wrote:
> On Wed, Jun 11, 2003 at 09:01:44PM +0200, PSI-Systems wrote:
> > Does this mean on your site that even with setting values in 
> > /proc/acpi/processor/CPU/performance you never get the processor clocking 
> > to the highest speed?
> 
> I made some more tests, and what previously confused me is that the state 
> is not properly initialized when the kernel module is loaded. The initial
> value given by /proc/acpi/processor/CPU/performance is always P0 
> (the maximum performance level), whatever the real processor state may be.
> 
> Later, modifiying this value has the expected behaviour, and
> allows me to alter the CPU available power. On a sample application,
> the elapsed user time ranges from 4.3s in P0 mode, to 11.3s in P5 mode.
> 
> Moreover, today on the LKML, I read this message stating that
> a free driver for enhanced speedstep has been written for
> the 2.5 kernel series :
> http://www.ussg.iu.edu/hypermail/linux/kernel/0306.1/1004.html
> 
> best wishes,
> -- 
> fabrice