I have heard of other intel cpu having a hardcoded lowest voltage. Mine is a duo T2500 and it has the same hardcode, mine at .95v. Since my cpu works at .95v for the next available higher frequency, I thought to just set the base frequency from the default lowest of 1,000,000 to 1,333,000 in that it aught to use the same power but yield a higher performance gain. So far so good.
Here is what I did to /etc/rc.local to set it for boot..
# By default this script does nothing.
echo 1333000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq
echo 1333000 > /sys/devices/system/cpu/cpu1/cpufreq/scaling_min_freq
exit 0
Workaround for a hardwired lowest voltage
-
- Administrator
- Posts: 346
- Joined: Wed 9. Jul 2008, 19:57 [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: Workaround for a hardwired lowest voltage
Heya. Great "hack".
But if you want to save even the latest microampere of power this solution is not the best.
With PHC you reduce the voltage the CPU gets, but you can not reduce the current (amperes) the CPU takes. And it takes more current the higher the frequency is.
How much the difference is one need to test on its own machine.
But if you want to save even the latest microampere of power this solution is not the best.
With PHC you reduce the voltage the CPU gets, but you can not reduce the current (amperes) the CPU takes. And it takes more current the higher the frequency is.
How much the difference is one need to test on its own machine.
-
- Posts: 11
- Joined: Tue 2. Dec 2008, 04:45 [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: Workaround for a hardwired lowest voltage
I am sure there is a trade off somewhat.
Doesn't the current really only significantly increase when there are actual calculations performed as opposed to just the clock ticking faster?
Doesn't the current really only significantly increase when there are actual calculations performed as opposed to just the clock ticking faster?
-
- Administrator
- Posts: 346
- Joined: Wed 9. Jul 2008, 19:57 [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: Workaround for a hardwired lowest voltage
I am not sure how much effective this is. Maybe you can get some results from powertop?
-
- Posts: 14
- Joined: Sat 27. Feb 2010, 00:06 [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: Workaround for a hardwired lowest voltage
Interesting, I was already thinking about overclocking to avoid wasting voltage at the lower frequency, but I will have to check the delta power consumption when the cpu is idle.
-
- Posts: 6
- Joined: Sat 29. May 2010, 03:25 [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: Workaround for a hardwired lowest voltage
Roughly, the Voltage has an influence on power dissipation with a power of ^2, therefore being the part where you can do the most savings, but the frequency still has a linear influence. What that means in your specific case can't be said, because there are some other factors, like how fast the CPU can dynamically switch frequencies etc. But if the voltage for 1000 and 1333mhz is the same, it should be able to change it in nearly zero time, as otherwise it first steps up voltage and waits(very little time for humans, but still time) until the voltage reaches its level, only then it can change the frequency to the next higher one.
For the theory of the power dissipation, see for example: http://en.wikipedia.org/wiki/CPU_power_dissipation
P=C*V^2*f
while you can change the capacitance C only at engineering/manufacturing time, you are able to do runtime savings by changing the voltage V or the frequency f...
For the theory of the power dissipation, see for example: http://en.wikipedia.org/wiki/CPU_power_dissipation
P=C*V^2*f
while you can change the capacitance C only at engineering/manufacturing time, you are able to do runtime savings by changing the voltage V or the frequency f...