On Fri, 30 Aug 2013 13:00:03 +0200 Holger Levsen <holger@layer-acht.org> wrote: > reassign 721310 src:linux > severity 721310 wishlist > retitle 721310 please enable ACPI_INITRD_TABLE_OVERRIDE > # <waldi> | h01ger: no. this is a firmware bug, which may be possible to > # workaround with acpi table overrides > # <h01ger> | waldi: so which package is providing that firmware? firmware- > # linux? > #< jcristau> | h01ger: the hardware > thanks > > > Hi Andreas, > > thanks for your installation-report, much appreciated! > > On Freitag, 30. August 2013, Andreas Glaeser wrote: > > [...] When the original CPU was replaced by an AMD Turion 64 > > TL-50, I found that this is not running in a stable manner and used up > > more energy than reasonable, because there were complaints at boot-time > > about missing ACPI-tables in the BIOS for Powernow-K8 frequency-scaling. > > So I built a kernel version with frequency-scaling disabled, that runs OK, > > so far, now in the second version already on an even newer Turion 64 TL-58 > > CPU. The upside of this is that the wattage is still very low, at a > > maximum of about 23 W. The downside is that the CPU runs at its minimum > > frequency, which results in unattractive latencies and low performance. > > When I configure the new 3.10 kernel-sources, I saw that there is an > > option to inject ACPI-tables from the initrd into the kernel instead of > > fetching them from the BIOS. So this is probably the way to unlock the > > CPU-potential with this thin-client hardware. The wattage will be about 10 > > W higher then, the CPU-speed should double at least and extra active > > cooling will be necessary. At this time I did not check yet, where to get > > those ACPI-tables from. I guess one could try to read them out of socket > > S1 notebook BIOSes, or get them directly from the technical specification > > from AMD. There was no further research about this yet. The following data > > show the system with the TL-50 CPU, that was present at installation-time. > > Current dmesg-output with newer kernel- and CPU-version is also attached. > > > cheers, > Holger At this time, there seem to be no ACPI-tables in both firmware-linux packages, but because these would be AMD-CPU-specific, I suppose, they could be included in firmware-linux-nonfree ? The help in the kernel-configuration says, ACPI_INITRD_TABLE_OVERRIDE does not do anything by default, so it can be safely enabled. But probably not many people do realy need this, except myself.
Attachment:
signature.asc
Description: PGP signature