[HowTo] Undervolting the AMD Turion X2 (and some others)

[exjinn]

Hi guys, I'm the author of TurionPowerControl. I saw in my web server log this discussion on ubuntu forums and so I got here to give you support if you wish.

Feel free to ask any questions about the program and its usage.



ps: I forgot to say that you can read the latest news and download the latest versione here: http://forum.notebookreview.com/hp-c...ng-hp-dv5.html

Hi blackshard thanks for your work.

If you look above the only issue I'm having is I'm not sure if it's actually doing what it's supposed to do.

I have a AMD Turion X2 Dual-Core Mobile RM-72 however the turion script says: Detected processor: Turion Ultra RM processor. Is this a mistake or something I setup improperly?

[mothersh1p]

Hi blackshard thanks for your work.

If you look above the only issue I'm having is I'm not sure if it's actually doing what it's supposed to do.

I have a AMD Turion X2 Dual-Core Mobile RM-72 however the turion script says: Detected processor: Turion Ultra RM processor. Is this a mistake or something I setup improperly?

It's a bug. The setup is correct. You could try TurionPowerControl -CM.

[blackshard]

Hi blackshard thanks for your work.

If you look above the only issue I'm having is I'm not sure if it's actually doing what it's supposed to do.

I have a AMD Turion X2 Dual-Core Mobile RM-72 however the turion script says: Detected processor: Turion Ultra RM processor. Is this a mistake or something I setup improperly?

Well, Turion Ultra RM processor is probably a minor bug. Try latest version (0.21) and check if it has been corrected.

However it won't affect the performance of the software, it will work fine anyway.

Most important is the fact you see some kind of overclocking with the turionuv script you ran.

It is a script made for a 2.2 ghz processor. It will set your processor to 2.2 ghz, but actually it won't go over 2.1 ghz because its FID multiplier is locked and can't go over 13 (instead in the script you see that FID of pstate0 is set to 13). All the programs of this world will tell you that your processor is working at 2.2 ghz, but if you do a simple benchmark, you will see that it has the same performances of 2.1 ghz. Also don't believe at all the cpufreq-* stats, they are not updated since the cpu scaler daemon starts, despite the fact cpufreq-info says that it is obtaining information with a hardware call.

Also notice that you can put all the switches on a single command line instead of calling the program many times. Also newer versions can have an external configuration file to recall settings.

[mothersh1p]

Well, Turion Ultra RM processor is probably a minor bug. Try latest version (0.21) and check if it has been corrected.

Hi blackshard,

thank you for your help! Although I think that your program is only made for 64 bit, I have found this switch to compile it for 32 bit:

c++ -D_FILE_OFFSET_BITS=64 TurionPowerControl.cpp cpuPrimitives.cpp >Processor.cpp config.cpp -o TurionPowerControl

Maybe it is useful to add it to your Makefile? (IF-THEN) Could it be that the turionuv-script doesn't work properly for 32 bit?

Although, in the sample output (using your program) from exjinn, I cannot see that pstate 0 has a FID of 13 after he ran turionuv. Could you clarify this?

[blackshard]

Hi blackshard,

thank you for your help! Although I have found that your program is only made for 64 bit, I have found this switch to compile it for 32 bit:

c++ -D_FILE_OFFSET_BITS=64 TurionPowerControl.cpp cpuPrimitives.cpp >Processor.cpp config.cpp -o TurionPowerControl

Maybe it is useful to add it to your Makefile? (IF-THEN) Could it be that the turionuv-script doesn't work properly for 32 bit?

Although, in the sample output (using your program) from exjinn, I cannot see that pstate 0 has a FID of 13 after he or she ran turionuv. Could you clarify this?

Hello mothersh1p, thank you for your notice about compiling on 32 bit platforms. Actually, as far as I remember, I once tried to compile it on a 32 bit fedora box with success. I'll try as soon as possible your solution (isn't there a ">" in the wrong place?) and will report it in the readme file.

exjinn said he has a processor running at 2.1 Ghz, it means that its maximum FID is 13 (just check the little formula in the readme file to calculate frequency with DID and FID given values).
All family 11h processors (Athlon QL, Turion RM/ZM processors) cap the overflowing values to the hardwired maximum value. In case of exjinn, the maximum FID value is 13, so you can set FID=14, 15 or even 20, the processor will accept the value but then it will internally cap that value to 13.

[mothersh1p]

Hello mothersh1p, thank you for your notice about compiling on 32 bit platforms. Actually, as far as I remember, I once tried to compile it on a 32 bit fedora box with success. I'll try as soon as possible your solution (isn't there a ">" in the wrong place?) and will report it in the readme file.

Hi Blackshard,

my apologize, the "greater"-sign is a typo. Just erase the ">" typo to make the cflag working. Would you mind to tell me

in my web server log this discussion on ubuntu forums

how can this be possible? Do you have an own website-crawler? Do you work for Google? Or do you plan to use such technology?

[blackshard]

Hi Blackshard,

my apologize, the "greater"-sign is a typo. Just erase the ">" typo to make the cflag working. Would you mind to tell me
how can this be possible? Do you have an own website-crawler? Do you work for Google? Or do you plan to use such technology?

Ehehe, nothing so exotic

I just told apache on my personal webserver to enable referrer logging

[mothersh1p]

I just told apache on my personal webserver to enable referrer logging

My apologize for my english. I'm a bit paranoid about new technology.

Here are the bits again in case you need it:

c++ -D_FILE_OFFSET_BITS=64 TurionPowerControl.cpp cpuPrimitives.cpp Processor.cpp config.cpp -o TurionPowerControl

Why didn't you use a Makefile? I guess it isn't possible to guess 32 bit boxes? I should try this oneliner, every time I boot into Linux I get this big output! XXX

[blackshard]

Mmmh, I just tried to compile on a 32 bit box running Ubuntu 10.04 LTS and compilation went good just using the compile.sh script included with the package.

I'd like to understand what is the problem related to 32bit cpuid instruction to let the program compile and run well even without the compilation flag.
I prefer not to give a makefile just because it is a bunch of .cpp files and a trivial shell script is just enough to compile the program.

[mothersh1p]

I'd like to understand what is the problem related to 32bit cpuid instruction to let the program compile and run well even without the compilation flag.
I prefer not to give a makefile just because it is a bunch of .cpp files and a trivial shell script is just enough to compile the program.

Maybe you want me to make your Makefile?