user undervolting

Message

mbar · Post by **mbar** » Fri Feb 24, 2006 12:40 pm

this is fcvking awesome!
My P-M 1,5 GHz undervolted from 1484 mV to 1212 mV (at 1500 MHz) and its temperature dropped from 85-88 C to 63-68 C

I no longer hear full speed fan noise

zhark · Post by **zhark** » Sun Feb 26, 2006 2:33 pm

soulwarrior wrote:I am too running this patch successfully

It's a Intel Pentium M 740 with the following values for the voltage table:

"956,956,956,956,956,956,956,828,732,700"

Greetings,
soulwarrior

I've also got the Pentium M 740 (1.73GHz):
Cpu Family: 6, Model: 13, Stepping 8, cpu MHz: 1729.341

I've enabled intel enhanced speedstep and cpu freq debugging in kernel, and i get this error message:
speedstep-centrino: Invalid control/status registers (1 - 1)
speedstep-centrino: <6>speedstep-centrino: invalid ACPI data
speedstep-centrino: <6>speedstep-centrino: found unsupported CPU with
Enhanced SpeedStep: send /proc/cpuinfo to Jeremy Fitzhardinge <jeremy@goop.org>
cpufreq-core: initialization failed
cpufreq-core: no CPU initialized for driver centrino
cpufreq-core: unregistering CPU 0

PS: This happens regardless of me applying the undervolt patch (in the howto) to speedstep-centrino.c

EDIT: Got it working!! Manually edited the original speedstep-centrino.c - Mine also does 533MHz!
See it here: http://zhark.bjorli.com/speedstep-centrino.c-zharkbb (This file is based on the 2.6.16-rc4 vanilla kernel)
(This file is modified to support Pentium M 740 1.73GHz, but can easily be altered to support others)

EDIT: Updated voltages after verification with mprime
Here are my voltages:
533MHz:700 800MHz:748 1733MHz: 956

-zhark

EASYdoor · Post by **EASYdoor** » Thu Mar 09, 2006 2:20 pm

damn,...i can't get this speedstep centrino code working, nomatter which options i try to inlcude/exclude to the kernel i always get acpi drivers loaded into the kernel, so i cannot manually control my centrino,...anyone else found a solution for this problem?

Code: Select all

cpufreq-info
cpufrequtils 0.4: cpufreq-info (C) Dominik Brodowski 2004
Report errors and bugs to linux@brodo.de, please.
analyzing CPU 0:
  no or unknown cpufreq driver is active on this CP

trying to load as a module,...

Code: Select all

FATAL: Error inserting speedstep_centrino (/lib/modules/2.6.16-rc3-nitro1/kernel/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.ko): No such device

And nomather wich kernel i run, same problem...
And I'm running Intel Pentium M 750 (2 Ghz), Acer TravelMate 8006

Code: Select all

processor	: 0
vendor_id	: GenuineIntel
cpu family	: 6
model		: 13
model name	: Intel(R) Pentium(R) M processor 2.00GHz
stepping	: 6
cpu MHz		: 598.289
cache size	: 2048 KB
fdiv_bug	: no
hlt_bug		: no
f00f_bug	: no
coma_bug	: no
fpu		: yes
fpu_exception	: yes
cpuid level	: 2
wp		: yes
flags		: fpu vme de pse tsc msr mce cx8 sep mtrr pge mca cmov pat clflush dts acpi mmx fxsr sse sse2 ss tm pbe est tm2
bogomips	: 1196.65

dmesg:

Code: Select all

* cpufreqd requires the kernel to be configured with CONFIG_CPU_FREQ

bdz · Post by **bdz** » Thu Mar 09, 2006 7:17 pm

As far as I know the centrino cpu-freq driver does not work correctly when compiled as a module. It has to be built into the kernel.
Here is how I have the CPU frequency scaling part of my kernel configured:

Code: Select all

Linux Kernel v2.6.15-gentoo-r6 Configuration
 ──────────────────────────────────────────────────────────────────────────────
  ┌───────────────────────── CPU Frequency scaling ─────────────────────────┐
  │  Arrow keys navigate the menu.  <Enter> selects submenus --->.          │
  │  Highlighted letters are hotkeys.  Pressing <Y> includes, <N> excludes, │
  │  <M> modularizes features.  Press <Esc><Esc> to exit, <?> for Help, </> │
  │  for Search.  Legend: [*] built-in  [ ] excluded  <M> module  < >       │
  │ ┌─────────────────────────────────────────────────────────────────────┐ │
  │ │[*] CPU Frequency scaling                                            │ │
  │ │[*]   Enable CPUfreq debugging                                       │ │
  │ │<*>   CPU frequency translation statistics                           │ │
  │ │[*]     CPU frequency translation statistics details                 │ │
  │ │      Default CPUFreq governor (performance)  --->                   │ │
  │ │---   'performance' governor                                         │ │
  │ │<*>   'powersave' governor                                           │ │
  │ │<*>   'userspace' governor for userspace frequency scaling           │ │
  │ │<*>   'ondemand' cpufreq policy governor                             │ │
  │ │<*>   'conservative' cpufreq governor                                │ │
  │ │---   CPUFreq processor drivers                                      │ │
  │ │< >   ACPI Processor P-States driver                                 │ │
  │ │< >   AMD Mobile K6-2/K6-3 PowerNow!                                 │ │
  │ │< >   AMD Mobile Athlon/Duron PowerNow!                              │ │
  │ │< >   AMD Opteron/Athlon64 PowerNow!                                 │ │
  │ │< >   Cyrix MediaGX/NatSemi Geode Suspend Modulation                 │ │
  │ │<*>   Intel Enhanced SpeedStep                                       │ │
  │ │[*]     Use ACPI tables to decode valid frequency/voltage pairs      │ │
  │ │[*]     Built-in tables for Banias CPUs                              │ │
  │ │< >   Intel Speedstep on ICH-M chipsets (ioport interface)           │ │
  │ │< >   Intel SpeedStep on 440BX/ZX/MX chipsets (SMI interface)        │ │
  │ │< >   Intel Pentium 4 clock modulation                               │ │
  │ │< >   nVidia nForce2 FSB changing                                    │ │
  │ │< >   Transmeta LongRun                                              │ │
  │ │< >   VIA Cyrix III Longhaul                                         │ │
  │ │---   shared options                                                 │ │
  │ │[ ]   /proc/acpi/processor/../performance interface (deprecated) (NEW│ │
  │ └─────────────────────────────────────────────────────────────────────┘ │
  ├─────────────────────────────────────────────────────────────────────────┤
  │                    <Select>    < Exit >    < Help >                     │
  └─────────────────────────────────────────────────────────────────────────┘

EASYdoor · Post by **EASYdoor** » Mon Mar 13, 2006 11:07 pm

finnaly got it working...extra 40 min of battery life with:

Code: Select all

scaling_voltages
600000 700

rovclock -c 60 -m 60

cool'n quiet ;) sweet

if anyone needs a modiyfied speedstep-centrino.c for Dothan 2.0 let me know

Elmo234 · Post by **Elmo234** » Wed Mar 15, 2006 10:52 pm

I just started coding a program that is used as a daemon and undervolts the athlon64 and centrino (Pentium-M and Celeron-M) cpu's.

I write the new vid's using /dev/cpu/*/msr

You can have a look at it here:

http://tuxamito.com.es/cpupw/

by the way... does anybody know how could I find if a Pentium-M (centrino) is using a 100 (400) or a 133 (533) Mhz bus?

Thanks

Muffe · Post by **Muffe** » Mon Mar 20, 2006 1:59 pm

Is there any way I can set the voltage below 700 mV? I have an Pentium-M 740 1,73 GHz, and have applied the patch as described in the WIKI, and is now able to set the voltage at 800 MHz down to 700 mV. Because I want to tune it as far down as possible, I tried to set it to 695 mV, but I was not able to do that.

Why does this lower limit exist? Is it any way I can bypass it or remove it? Thanks.

mbar · Post by **mbar** » Mon Mar 20, 2006 2:43 pm

It's an MSR register limitation. You cannot do anything about it.

Keruskerfuerst · Post by **Keruskerfuerst** » Mon Mar 20, 2006 7:06 pm

Muffe wrote:Is there any way I can set the voltage below 700 mV? I have an Pentium-M 740 1,73 GHz, and have applied the patch as described in the WIKI, and is now able to set the voltage at 800 MHz down to 700 mV. Because I want to tune it as far down as possible, I tried to set it to 695 mV, but I was not able to do that.

Why does this lower limit exist? Is it any way I can bypass it or remove it? Thanks.

There is also a technical limitation.
Depends on the endowment.
This results in a level, where the transistors do switch from "low" to "high" and vice versa.

rschwarze · Post by **rschwarze** » Sun Apr 02, 2006 3:56 pm

Hey guys!

We have done a new version of the undervolting patch. We now call it "Linux PHC" and it is hosted on sourceforge.

The Project Website is: http://linux-phc.sourceforge.net

If you like, please try out the new version and report any problems or other feedback into this thread.

Thank you very much,

Roman

jwj · Post by **jwj** » Sun Apr 02, 2006 6:57 pm

Hi rschwarze,

Thanks for your phc patch. I applied it to linux-2.6.16-ck2:

Code: Select all

bash # echo "1862000:1036,1862000:1036,1862000:1036,1862000:1036,1862000:1036,1862000:1036,1596000:956,1330000:876,1064000:780,798000:700" > /sys/devices/system/cpu/cpu0/cpufreq/op_points_table

Code: Select all

bash # cat /sys/devices/system/cpu/cpu0/cpufreq/op_points_table
1862000:1036,1862000:1356,1862000:1356,1862000:1356,1862000:1356,1862000:1356,1596000:956,1330000:876,1064000:780,798000:700

Only the first 1862000 entry seems to be changed, dont know, if this is a problem, but it seems my fan is running more than with the bdz patch.

rschwarze · Post by **rschwarze** » Sun Apr 02, 2006 9:14 pm

you may also want to check the values in /sys/devices/system/cpu/cpu0/cpufreq/voltage_table, because that interface is still available.

Can you tell me if the entrys in the voltage_table look different then with the old patch?

Thank you for testing!

jwj · Post by **jwj** » Sun Apr 02, 2006 9:36 pm

with the phc patch:

Code: Select all

bash # cat /sys/devices/system/cpu/cpu0/cpufreq/voltage_table
1036,1356,1356,1356,1356,1356,956,876,780,700

bash # echo "1036,1036,1036,1036,1036,1036,956,876,780,700" > /sys/devices/system/cpu/cpu0/cpufreq/voltage_table

bash # cat /sys/devices/system/cpu/cpu0/cpufreq/voltage_table
1036,1036,1036,1036,1036,1036,956,876,780,700

bash # cat /sys/devices/system/cpu/cpu0/cpufreq/op_points_table
1862000:1036,1862000:1036,1862000:1036,1862000:1036,1862000:1036,1862000:1036,1596000:956,1330000:876,1064000:780,798000:700

So it seems, that after setting the values in voltage_table they also show up correct in op_points_table.

Output of bdz patched 2.6.15 kernel:

Code: Select all

bash # echo "1036,1036,1036,1036,1036,1036,940,876,780,700" > /sys/devices/system/cpu/cpu0/cpufreq/voltage_table

bash # cat /sys/devices/system/cpu/cpu0/cpufreq/voltage_table
1036,1036,1036,1036,1036,1036,940,876,780,700

Hope this helps.

rschwarze · Post by **rschwarze** » Sun Apr 02, 2006 10:12 pm

Yes, thats what i thought...

I don't have this problem with my notebook, because i don't have multiple op-points:

Code: Select all

bash $ cat /sys/devices/system/cpu/cpu0/cpufreq/op_points_table
600000:700,800000:716,1000000:780,1200000:860,1400000:956

Code: Select all

bash $ cat /sys/devices/system/cpu/cpu0/cpufreq/voltage_table  
700,716,780,860,956

bdz will have a look at it, as soon as he has time. I hope we can solve this soon.

Thanks again for your feedback!

bdz · Post by **bdz** » Tue Apr 04, 2006 9:44 pm

jwj wrote:Hi rschwarze,

Thanks for your phc patch. I applied it to linux-2.6.16-ck2:
Code: Select all
bash # echo "1862000:1036,1862000:1036,1862000:1036,1862000:1036,1862000:1036,1862000:1036,1596000:956,1330000:876,1064000:780,798000:700" > /sys/devices/system/cpu/cpu0/cpufreq/op_points_table
Code: Select all
bash # cat /sys/devices/system/cpu/cpu0/cpufreq/op_points_table
1862000:1036,1862000:1356,1862000:1356,1862000:1356,1862000:1356,1862000:1356,1596000:956,1330000:876,1064000:780,798000:700
Only the first 1862000 entry seems to be changed, dont know, if this is a problem, but it seems my fan is running more than with the bdz patch.

Hi jwj!

I have reproduced your problem on my laptop where I also have these duplicated operating points when using ACPI table. I did not noticed the problem before because I'm usualy using built-in table.
I have made the correction in the patch and tested it on my laptop. Unfortunalty Sourceforge is currently down so I can't upload the new release right now. I will keep you informed as soon as I'm able do do it.

Edit:
Sourceforge is back up. I have uploaded the new release (0.2.2) :
https://sourceforge.net/project/showfil ... _id=161063

jwj · Post by **jwj** » Tue Apr 04, 2006 11:20 pm

Thank you very much bdz, the 0.2.2 release fixes the problem here.

-jwj

Egalus · Post by **Egalus** » Fri Apr 07, 2006 12:09 pm

Hi,

first of all thank you for your great work. I lack the c-knowledge to contribute in any other matter than testing and making comments.

At the moment I am using the old version of the patch (not the linux-phc one) with kernel 2.6.15 with a pinmodded 1.7GHz P-M which now is running at 2.26GHz max-speed.
With this old patch every tool reports the cpu to be running at the desired 533 FSB.

As the 2.6.16 kernel is out and the changelog claims that some work regarding powerrsaving is done (and 2.6.15 while idling still consumes about 10% more juce than winxp) I want to change over to this new kernel.
I use a vanilla 2.6.16.1 kernel and the 0.2.2 linux-phc patch for 2.6.16 with the default settings suggested in the Patching.txt file.

So far so good, I get the needed devices in /sys/ , but somehow all programms that show cpu-speed now show a 600-1700 Mhz P-M on a 400 MHz FSB.
This leads to some questions:
a) is the cpu now running at 400 or 533 fsb (as the mod is a hardware mod I would be surprised if it's running at 400)
b) how do I find out at which speed the cpu really is running
c) if this is just a coincidence of the builtin tables for dothan for detecting the correct voltages and by that translating the 1.7 P-M (which is pinmodded to be a 2.26 P-M) back to a normal 1.7 P-M is going to be changed in future releases?

And one comment on the naming of the tables.
Banias and Dothan really are processors,
but Sonoma actually is a chipset and you surely mean 533 FSB Dothans with the name.
There is already so much confusion about celeron, banias, dothans, I guess naming this table "sonoma" might add to confusion.

Regards,
Norbert

brihall · Post by **brihall** » Sat Apr 08, 2006 5:30 pm

I have a laptop with a Pentium-III mobile CPU (_not_ centrino!). Speedstep (normal, not enhanced) is enabled and working fine, giving me two frequencies, 800 and 1200 Mhz. Anyone have a patch or tips on how to undervolt a P3-M CPU with speedstep?

00:00.0 Host bridge: Intel Corporation 82815 815 Chipset Host Bridge and Memory Controller Hub (rev 04)
00:01.0 PCI bridge: Intel Corporation 82815 815 Chipset AGP Bridge (rev 04)
00:1e.0 PCI bridge: Intel Corporation 82801 Mobile PCI Bridge (rev 03)
00:1f.0 ISA bridge: Intel Corporation 82801BAM ISA Bridge (LPC) (rev 03)
00:1f.1 IDE interface: Intel Corporation 82801BAM IDE U100 (rev 03)
00:1f.2 USB Controller: Intel Corporation 82801BA/BAM USB (Hub #1) (rev 03)

processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 11
model name : Intel(R) Pentium(R) III Mobile CPU 1200MHz
stepping : 1
cpu MHz : 798.000
cache size : 512 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 2
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 mmx fxsr sse
bogomips : 1596.39

rschwarze · Post by **rschwarze** » Mon Apr 10, 2006 9:09 pm

@egalus:

recently, someone wrote me that he used the patch for an overclocked dothan in the following way. its just a workaround, but maybe you wanna try it:

I have a Dothan 735 (1.7Ghz with 100Mhz bus) overclocked to
2.2Ghz by changing the fsb to 133Mhz. A pristine 2.6.16 kernel
will correctly display "2267000" as the first frequency in
/sys/.../cpufreq/scaling_available_frequencies but after
applying the PHC patch it displays 170000 (the cpu still
runs at 2.2Ghz).

I had a look at speedstep-centrino.c and if I removed these
lines (line 570 - 575):

if ((centrino_model[cpu]) && (centrino_model[cpu]->base_freq != 0))
{
msr = (msr >>

& 0xff;
return msr * centrino_model[cpu]->base_freq;
}

/sys/.../cpufreq/scaling_available_frequencies will again show
"2267000" as the first frequency.

131486 · Post by **131486** » Tue Apr 11, 2006 4:32 pm

hi,
i'm a new gentoo user, i would learn how to modify my speedstep-centrino.c.

i'm currently using centrino driver:

Code: Select all

# cpufreq-info
cpufrequtils 0.4: cpufreq-info (C) Dominik Brodowski 2004
Report errors and bugs to linux@brodo.de, please.
analyzing CPU 0:
  driver: centrino
  CPUs which need to switch frequency at the same time: 0
  hardware limits: 800 MHz - 1.60 GHz
  available frequency steps: 1.60 GHz, 1.33 GHz, 1.07 GHz, 800 MHz
  available cpufreq governors: conservative, ondemand, powersave, performance
  current policy: frequency should be within 800 MHz and 1.60 GHz.
                  The governor "ondemand" may decide which speed to use
                  within this range.
  current CPU frequency is 800 MHz (asserted by call to hardware).

How to get minimum frequency of 533mhz?it is possible to go under?
which speedstep-centrino.c i have to modify? /usr/src/linux-2.6.16-gentoo-r1/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c ?
To apply settings, i have to recompile and reboot?

rschwarze · Post by **rschwarze** » Tue Apr 11, 2006 5:20 pm

1. Dowload the linux-phc patch from http://linux-phc.sourceforge.net/
2. Apply the patch according to the documentation provided with the patch
3. use "make menuconfig" to configure your kernel and configure the patch to use "build in tables" for your processor
4. Make any changes to the speedstep-centrino.c file in the kernel directory that you like (to use undervolting, you dont have to change anything)
5. save your .config
6. "make clean"
7. restore yout .config
8. "make && make modules_install"
9. copy the kernel to /boot and change your grub or lilo settings accordingly

131486 · Post by **131486** » Tue Apr 11, 2006 5:25 pm

rschwarze wrote:1. Dowload the linux-phc patch from http://linux-phc.sourceforge.net/
4. Make any changes to the speedstep-centrino.c file in the kernel directory that you like (to use undervolting, you dont have to change anything)

well, i really don't know which section i have to modify in order to get 533mhz and to get lower vids...

thanks alot!

131486 · Post by **131486** » Wed Apr 12, 2006 12:39 am

got it!

Code: Select all

static struct cpufreq_frequency_table sonoma_1596[] =
{
	OPEX( 533, 133,  700,  700,  700,  700),  /* 533mhz  */
	OPEX( 798, 133,  988,  988,  988,  988),
	OPEX(1064, 133, 1116, 1111, 1084, 1079),
	OPEX(1330, 133, 1244, 1233, 1180, 1169),
	OPEX(1596, 133, 1356, 1356, 1260, 1260),
	{ .frequency = CPUFREQ_TABLE_END }
};

however i can't go under 0.7v as voltage...may be i have to modify something else?

rschwarze · Post by **rschwarze** » Wed Apr 12, 2006 2:45 am

no, its not possible to run below 700mv. its a hardware limitation.

however, are you sure its running at 533? you can write whatever you want into the speedstep-centino.c and cpufreq will show it as a possible frequency. to be sure, you should use some benchmark to see that its really running slower at that frequency. for example you can just use the benchmark of mprime.

good to now that everything else works just fine for you!

131486 · Post by **131486** » Wed Apr 12, 2006 2:49 am

rschwarze wrote:no, its not possible to run below 700mv. its a hardware limitation.

however, are you sure its running at 533? you can write whatever you want into the speedstep-centino.c and cpufreq will show it as a possible frequency. to be sure, you should use some benchmark to see that its really running slower at that frequency. for example you can just use the benchmark of mprime.

good to now that everything else works just fine for you!

not sure at all...running kpowersave or cpuinfo (i suppose they use cpufrequtils or something in /proc) give me that value...
what should i do running mprime?
In your opition i can run more cool at 533mhz@0.7 vs 800@0.7?or it is just the same because of the same voltage?

user undervolting

Go below 700 mV

Re: Go below 700 mV