PST tables and Sonys Bios

PST tables and Sonys Bios

[Lars Gemeinhardt]

Hi developer and user,

Sony Bios (and the PST tables) maybe don't like the powernow feature, 
because yesterday I have updated my FX705 from Bios (R0117K5) to the 
latest one (R0121K5) and then I dump all PST tables (with powernow 
infos) and found that they do not fix the "PST:26" for my CPU (all VIDs 
are 1.3V), no they change the correct (?) "PST:27" (maybe for the 1600+) 
from:

 PST:27 (@0x401906ce)
  cpuid: 0x780    fsb: 100        maxFID: 0x16    startvid: 0x11
  num of p states in this table: 6
    FID: 0x4 (5.0x [500MHz])    VID: 0x19 (1.050V)
    FID: 0x6 (6.0x [600MHz])    VID: 0x19 (1.050V)
    FID: 0x8 (7.0x [700MHz])    VID: 0x19 (1.050V)
    FID: 0xc (9.0x [900MHz])    VID: 0x19 (1.050V)
    FID: 0x0 (11.0x [1100MHz])  VID: 0x15 (1.150V)
    FID: 0x16 (14.0x [1400MHz]) VID: 0xe (1.300V)

to:

 PST:27 (@0x401906be)
  cpuid: 0x780    fsb: 100        maxFID: 0x16    startvid: 0xe
  num of p states in this table: 6
    FID: 0x4 (5.0x [500MHz])    VID: 0xe (1.300V)
    FID: 0x6 (6.0x [600MHz])    VID: 0xe (1.300V)
    FID: 0xa (8.0x [800MHz])    VID: 0xe (1.300V)
    FID: 0xe (10.0x [1000MHz])  VID: 0xe (1.300V)
    FID: 0x2 (12.0x [1200MHz])  VID: 0xe (1.300V)
    FID: 0x16 (14.0x [1400MHz]) VID: 0xe (1.300V)

Now the VIDs for this cpu are also all 1.3V :-(

Is there a chance to include a fix in the powernow-k7 code "for broken 
PST tables" or should everybody make it's own patch like "Luigi Belli" 
in "Acer Aspire 1300 buggy BIOS - hardcoded patch" 
(http://www.linux.org.uk/mailman/private/cpufreq/2003-June/002171.html)?

Ciao Lars

Re: PST tables and Sonys Bios

[Carl Thompson]

My Sony Vaios (I no longer have them) both were in the FXA series (same as
the FX series) and both had completely correct Powernow tables for the
Mobile Athlon 4 part for both of the BIOSs you mentioned.

Fact 1: the PSTs weren't broken at all in any BIOS for my Mobile Athlon 4
CPUs.

Fact 2: the PSTs for some models of Mobile Athlon XP started off kinda
"broken" and got more "broken" with the lastest BIOS.

Fact 3: that BIOS update supposedly fixes several lockup problems due to
"system conflicts" according to the docs.

Conclusion: my guess is that Sony purposely bumped up the voltage for _some_
CPUs because they were having system stability problems related to voltage
scaling with them.  It's possible that the PSTs are like this to actually
get your machine to run properly.

You might want to consider that the PST as it is in the BIOS might be
required to keep your machine from locking up occasionally.  Also rememeber
that with those PSTs CPUFreq will still scale the CPU speed and save you
power, it just won't scale voltage too.  Maybe getting the extra few
minutes of battery power that scaling the voltage might give you isn't
worth the hassle?

Of course, on those Sonys, the battery life is so crappy that you may be
willing to put up with some system instability to go longer...

It's also possible that the problem may be Windows specific and Linux /
CPUFreq work fine scaling the voltage of your processor.

Just some food for though,
Carl Thompson

Quoting Lars Gemeinhardt <lars.gemeinhardt@searchbroker.de>:

> Hi developer and user,
>
> Sony Bios (and the PST tables) maybe don't like the powernow feature,
> because yesterday I have updated my FX705 from Bios (R0117K5) to the
> latest one (R0121K5) and then I dump all PST tables (with powernow
> infos) and found that they do not fix the "PST:26" for my CPU (all VIDs
> are 1.3V), no they change the correct (?) "PST:27" (maybe for the 1600+)
> from:
>
>  PST:27 (@0x401906ce)
>   cpuid: 0x780    fsb: 100        maxFID: 0x16    startvid: 0x11
>   num of p states in this table: 6
>     FID: 0x4 (5.0x [500MHz])    VID: 0x19 (1.050V)
>     FID: 0x6 (6.0x [600MHz])    VID: 0x19 (1.050V)
>     FID: 0x8 (7.0x [700MHz])    VID: 0x19 (1.050V)
>     FID: 0xc (9.0x [900MHz])    VID: 0x19 (1.050V)
>     FID: 0x0 (11.0x [1100MHz])  VID: 0x15 (1.150V)
>     FID: 0x16 (14.0x [1400MHz]) VID: 0xe (1.300V)
>
> to:
>
>  PST:27 (@0x401906be)
>   cpuid: 0x780    fsb: 100        maxFID: 0x16    startvid: 0xe
>   num of p states in this table: 6
>     FID: 0x4 (5.0x [500MHz])    VID: 0xe (1.300V)
>     FID: 0x6 (6.0x [600MHz])    VID: 0xe (1.300V)
>     FID: 0xa (8.0x [800MHz])    VID: 0xe (1.300V)
>     FID: 0xe (10.0x [1000MHz])  VID: 0xe (1.300V)
>     FID: 0x2 (12.0x [1200MHz])  VID: 0xe (1.300V)
>     FID: 0x16 (14.0x [1400MHz]) VID: 0xe (1.300V)
>
> Now the VIDs for this cpu are also all 1.3V :-(
>
> Is there a chance to include a fix in the powernow-k7 code "for broken
> PST tables" or should everybody make it's own patch like "Luigi Belli"
> in "Acer Aspire 1300 buggy BIOS - hardcoded patch"
> (http://www.linux.org.uk/mailman/private/cpufreq/2003-June/002171.html)?
>
> Ciao Lars
>
>
> _______________________________________________
> Cpufreq mailing list
> Cpufreq@www.linux.org.uk
> http://www.linux.org.uk/mailman/listinfo/cpufreq
>
>

Re: PST tables and Sonys Bios

[Johannes von Loewis]

On Thu, Jul 17, 2003 at 09:15:08AM -0700, Carl Thompson wrote:
> Fact 3: that BIOS update supposedly fixes several lockup problems due to
> "system conflicts" according to the docs.
> 
> Conclusion: my guess is that Sony purposely bumped up the voltage for _some_
> CPUs because they were having system stability problems related to voltage
> scaling with them.  It's possible that the PSTs are like this to actually
> get your machine to run properly.
>
> You might want to consider that the PST as it is in the BIOS might be
> required to keep your machine from locking up occasionally.  Also rememeber

My machine (Sony Vaio FX805, Linux 2.4.2[01]+acpi+swsusp+cpufreq) locks
up occasionally. The BIOS PST is a "new one" with the voltage of 1.3V
for all frequencies. I have not yet found out, what the reason(s) for
the lock-ups is/are. Is it a problem with the standard kernel, a problem
with one of the patches, or a problem with the softmodem driver
hsflinmodem-5.03.whatever? Unfortunately, the BIOS PST is not sufficient
to keep the machine from locking up;-)

> that with those PSTs CPUFreq will still scale the CPU speed and save you
> power, it just won't scale voltage too.  Maybe getting the extra few
> minutes of battery power that scaling the voltage might give you isn't
> worth the hassle?

On my laptop the effect of scaling only the frequency is not sufficient
to keep the CPU fan from occasionally turning on. I have the hope to get
the machine quiet (for low-load situations as text editing or mp3
playing) by scaling the voltage too. 

Therefore it seems desirable to me to have a means to override the BIOS
PST in order to test upto what voltage the machine does not lock up much
more often than with 1.3V.

Regards

Johannes

Re: PST tables and Sonys Bios

[Carl Thompson]

Quoting Johannes von Loewis <johannes@v.loewis.de>:

> ...

> My machine (Sony Vaio FX805, Linux 2.4.2[01]+acpi+swsusp+cpufreq) locks
> up occasionally. The BIOS PST is a "new one" with the voltage of 1.3V
> for all frequencies. I have not yet found out, what the reason(s) for
> the lock-ups is/are. Is it a problem with the standard kernel, a problem
> with one of the patches, or a problem with the softmodem driver
> hsflinmodem-5.03.whatever? Unfortunately, the BIOS PST is not sufficient
> to keep the machine from locking up;-)

My experience with the Sonys is that their fans alone are insufficient to
maintain an equilibrium under high load and they overheat _very_ quickly. 
On these machines you _must_ run a program to monitor CPU temperature and
scale/throttle the CPU down when it gets too hot or the system _will_ lock
up.  (My cpuspeed daemon does that as well as automatically adjust the the
speed depending on the demand on the processor.  After adding this to my
program I never experienced a lock up on my Sonys again.)

Do your lock ups coincide with heavy CPU use?  I ran exactly what you are
running (including the modem driver) and did not experience any lock ups
except for heat related ones.

The only other issue I had is that rarely ACPI would "freak out" when
battery power was extremely low (<2% IIRC).  And I only experienced that
problem when running multiple batteries.

> ...

> On my laptop the effect of scaling only the frequency is not sufficient
> to keep the CPU fan from occasionally turning on. I have the hope to get
> the machine quiet (for low-load situations as text editing or mp3
> playing) by scaling the voltage too.
>
> Therefore it seems desirable to me to have a means to override the BIOS
> PST in order to test upto what voltage the machine does not lock up much
> more often than with 1.3V.

In my experience on the Sonys (and most every other modern desktop
replacement laptop) even with full CPU scaling and ACPI C2 if you are doing
anything at all the fan will be on a good portion of the time even if the
CPU at at minimum speed and voltage.  Lower powered CPUs are better at
keeping the fan off; my Mobile Duron based Sony had the fan on much less
than my Mobile Athlon 4 based Sonys.  Playing MP3s is probably enough to
keep the fan running constantly on your Mobile Athlon XP machine.  Be happy
though-- the fans on the Sonys are quiter than most and _much_ quiter than
some (Compaq Presarios are horribly noisy).

Yes, I'm kind of a laptop junkie and over the last 3 years I've owned 7! 
Best: my current eMachines M5305 Widscreen.  Worst: KDS Valiant 6480iPTD-
actually fell apart from 6 months of normal use.

> Regards
>
> Johannes

Carl Thompson