Re: PROBLEM: system clock slow on Athlon AMD64 since 2.6.21

[Mikael Pettersson]

On Sun, 10 Jun 2007 12:02:00 -0600, Robert Hancock wrote:
> Mikael Pettersson wrote:
> > On Fri, 08 Jun 2007 10:14:03 +0200, Gerard H. Pille wrote:
> >> [1.] One line summary of the problem:
> >> Since I switched from 2.6.20 to 2.6.21 on my Athlon AMD64 laptop, the system 
> >> time is slow - about 1' on 15'.
> > 
> > According to your system description it seems that you have a
> > Targa Visionary laptop with a VIA chipset and a Mobile Athlon64.
> > If so, then you probably have the same problem I reported some time
> > ago: see <http://marc.info/?l=linux-kernel&m=117681226421346&w=2>
> > and the followup messages. The conclusion was that the chipset's
> > ACPI PM timer slows down when the CPU is in C2.
> > 
> > The workaround is to boot with processor.max_cstate=1.
> 
> Which, if true, is extremely bizarre, as the whole point and reason for 
> existence of the ACPI PM timer is that it NOT do this, and remain at the 
> same rate regardless of CPU power management changes..

If you read the thread you'll find another bizarre fact: my laptop
only saw acpi_pm lose time when running a 64-bit kernel. With the
32-bit kernel acpi_pm worked fine in C2.

The C1 workaround is what the ACPI folks came up with. I was reluctant
to use it at first, but haven't really noticed any negative effects.

/Mikael
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/