net6501 Speedstep/cpufreq

View: New views
6 Messages — Rating Filter:   Alert me  

net6501 Speedstep/cpufreq

by Ben Kochie-2 :: Rate this Message:

| View Threaded | Show Only this Message

I'm testing my net6501-70 boxes right now with Vyatta Linux
(2.6.37-1-586-vyatta)

The cpufreq modules don't seem to be loading properly.  This is likely
because it can't find the ACPI table from the BIOS to setup the
correct modes for the Atom E680.  I tried loading some of the modules
manually, but they don't seem to work.

The CPU runs at about 115F which isn't too bad, but it would be nice
if speestep worked.  Has anyone had any luck with this?  Maybe it's
possible to load the required tables from userspace.

# uname -a
Linux vyatta 2.6.37-1-586-vyatta #1 SMP Thu Jul 7 22:01:14 PDT 2011
i686 GNU/Linux

# lsmod | grep cpu
acpi_cpufreq            4411  0
processor              21816  1 acpi_cpufreq
mperf                    907  1 acpi_cpufreq
cpufreq_userspace       1392  0
cpufreq_stats           1950  0
cpufreq_powersave        614  0
cpufreq_ondemand        6180  0
freq_table              1679  3 acpi_cpufreq,cpufreq_stats,cpufreq_ondemand
cpufreq_conservative     6182  0

# cpufreq-info
cpufrequtils 007: cpufreq-info (C) Dominik Brodowski 2004-2009
Report errors and bugs to cpufreq@..., please.
analyzing CPU 0:
  no or unknown cpufreq driver is active on this CPU
  maximum transition latency: 0.00 ms.
analyzing CPU 1:
  no or unknown cpufreq driver is active on this CPU
  maximum transition latency: 0.00 ms.


# dmesg | grep -i cpu
[    0.000000] x86 PAT enabled: cpu 0, old 0x7040600070406, new 0x7010600070106
[    0.000000] Processor #0 (Bootup-CPU)
[    0.000000] SMP: Allowing 2 CPUs, 0 hotplug CPUs
[    0.000000] setup_percpu: NR_CPUS:32 nr_cpumask_bits:32
nr_cpu_ids:2 nr_node_ids:1
[    0.000000] PERCPU: Embedded 12 pages/cpu @f5000000 s25088 r0 d24064 u2097152
[    0.000000] pcpu-alloc: s25088 r0 d24064 u2097152 alloc=1*4194304
[    0.000000] pcpu-alloc: [0] 0 1
[    0.000000] Initializing CPU#0
[    0.000000] SLUB: Genslabs=15, HWalign=64, Order=0-3, MinObjects=0,
CPUs=2, Nodes=1
[    0.000000]  RCU-based detection of stalled CPUs is disabled.
[    0.000000] CPU 0 irqstacks, hard=f4c08000 soft=f4c0a000
[    0.010790] CPU: Physical Processor ID: 0
[    0.010803] CPU: Processor Core ID: 0
[    0.010817] mce: CPU supports 5 MCE banks
[    0.010849] CPU0: Thermal monitoring enabled (TM1)
[    0.114604] CPU0: Genuine Intel(R) CPU        @ 1.60GHz stepping 01
[    0.231080] CPU 1 irqstacks, hard=f4c62000 soft=f4c64000
[    0.020000] Initializing CPU#1
[    0.400054] Brought up 2 CPUs
[    3.474210] cpuidle: using governor ladder
[    3.474210] cpuidle: using governor menu
_______________________________________________
Soekris-tech mailing list
Soekris-tech@...
http://lists.soekris.com/mailman/listinfo/soekris-tech

Re: net6501 Speedstep/cpufreq

by Chris Boot :: Rate this Message:

| View Threaded | Show Only this Message


On 17 Mar 2012, at 19:22, Ben Kochie <superq@...> wrote:

> I'm testing my net6501-70 boxes right now with Vyatta Linux
> (2.6.37-1-586-vyatta)
>
> The cpufreq modules don't seem to be loading properly.  This is likely
> because it can't find the ACPI table from the BIOS to setup the
> correct modes for the Atom E680.  I tried loading some of the modules
> manually, but they don't seem to work.
>
> The CPU runs at about 115F which isn't too bad, but it would be nice
> if speestep worked.  Has anyone had any luck with this?  Maybe it's
> possible to load the required tables from userspace.

[snip]

The 6501 BIOS has no ACPI and there are no plans to implement it. Without ACPI you can't get cpufreq on these newer processors. Basically, you can't use cpufreq scaling on the 6501 and won't be able to for the foreseeable future.

HTH,
Chris

--
Chris Boot
bootc@...

_______________________________________________
Soekris-tech mailing list
Soekris-tech@...
http://lists.soekris.com/mailman/listinfo/soekris-tech

Re: net6501 Speedstep/cpufreq

by Stanislav Meduna :: Rate this Message:

| View Threaded | Show Only this Message

On 18.03.2012 13:03, Chris Boot wrote:

> The 6501 BIOS has no ACPI and there are no plans to implement
> it. Without ACPI you can't get cpufreq on these newer processors.
> Basically, you can't use cpufreq scaling on the 6501 and won't
> be able to for the foreseeable future.

You might try your luck in some of the linux mailing lists.
As long as only the cpufreq is requested this should not
be so hard to implement - basically taking the old
speedstep-centrino.c driver and dig the correct values
of the frequency/voltage table from the CPU manual.

If you find out how to contact this guy
http://software.intel.com/en-us/forums/showthread.php?t=61309
who tried this with another processor 3 years ago, maybe
he is able to give some hints.

Regards
--
                                  Stano
_______________________________________________
Soekris-tech mailing list
Soekris-tech@...
http://lists.soekris.com/mailman/listinfo/soekris-tech

Re: net6501 Speedstep/cpufreq

by Soren Kristensen :: Rate this Message:

| View Threaded | Show Only this Message

Hi,

Chris Boot wrote:

>
> On 17 Mar 2012, at 19:22, Ben Kochie<superq@...>  wrote:
>
>> I'm testing my net6501-70 boxes right now with Vyatta Linux
>> (2.6.37-1-586-vyatta)
>>
>> The cpufreq modules don't seem to be loading properly.  This is
>> likely because it can't find the ACPI table from the BIOS to setup
>> the correct modes for the Atom E680.  I tried loading some of the
>> modules manually, but they don't seem to work.
>>
>> The CPU runs at about 115F which isn't too bad, but it would be
>> nice if speestep worked.  Has anyone had any luck with this?  Maybe
>> it's possible to load the required tables from userspace.
>
> [snip]
>
> The 6501 BIOS has no ACPI and there are no plans to implement it.
> Without ACPI you can't get cpufreq on these newer processors.
> Basically, you can't use cpufreq scaling on the 6501 and won't be
> able to for the foreseeable future.

Although it's not highest priority, I am looking into implementing ACPI,
it's not that complicated as long as it's just the tables and not the
advanced power management....

But despite having all kind of docs on the Atom E6xx series, including
BIOS writers guide under RSNDA, I haven't managed yet to get the P-State
tables....

And there isn't that much power to save anyway, the Atom core itself is
using less power than the memory and io controllers....


Best Regards,


Soren Kristensen

CEO & Chief Engineer
Soekris Engineering, Inc.
_______________________________________________
Soekris-tech mailing list
Soekris-tech@...
http://lists.soekris.com/mailman/listinfo/soekris-tech

Re: net6501 Speedstep/cpufreq

by Stanislav Meduna :: Rate this Message:

| View Threaded | Show Only this Message

On 18.03.2012 14:07, Soren Kristensen wrote:

> But despite having all kind of docs on the Atom E6xx series, including
> BIOS writers guide under RSNDA, I haven't managed yet to get the P-State
> tables....

Wow, you are right - I digged through nearly everything and
did not found anything usable. When did this Intel policy
change?

Anyway, reading and saving the table from some existing
ACPI-enabled BIOS via a published interface should not count
as anything illegal, does it? - in that case the acpi-cpufreq
log messages would be also illegal... At least for a person
that did not sign any NDA.

> And there isn't that much power to save anyway, the Atom core itself is
> using less power than the memory and io controllers....

OK, then it probably does not really make sense anyway.

Regards
--
                                Stano
_______________________________________________
Soekris-tech mailing list
Soekris-tech@...
http://lists.soekris.com/mailman/listinfo/soekris-tech

Re: net6501 Speedstep/cpufreq

by Ben Kochie-2 :: Rate this Message:

| View Threaded | Show Only this Message

I wonder if the tables are in coreboot. I know some people who might be able to help with this.

On Mar 18, 2012 12:36 PM, "Stanislav Meduna" <stano@...> wrote:
On 18.03.2012 14:07, Soren Kristensen wrote:

> But despite having all kind of docs on the Atom E6xx series, including
> BIOS writers guide under RSNDA, I haven't managed yet to get the P-State
> tables....

Wow, you are right - I digged through nearly everything and
did not found anything usable. When did this Intel policy
change?

Anyway, reading and saving the table from some existing
ACPI-enabled BIOS via a published interface should not count
as anything illegal, does it? - in that case the acpi-cpufreq
log messages would be also illegal... At least for a person
that did not sign any NDA.

> And there isn't that much power to save anyway, the Atom core itself is
> using less power than the memory and io controllers....

OK, then it probably does not really make sense anyway.

Regards
--
                               Stano
_______________________________________________
Soekris-tech mailing list
Soekris-tech@...
http://lists.soekris.com/mailman/listinfo/soekris-tech

_______________________________________________
Soekris-tech mailing list
Soekris-tech@...
http://lists.soekris.com/mailman/listinfo/soekris-tech