Today I've tried undervolting my Intel Core2Duo T9400 with the help of the phc-intel tool which allows to dynamically change the CPU voltage in Linux. It was something I was going to do long ago, but it turned out that a script that is used to check voltages supported by the CPU requires userspace cpu-freq governor, for which I had to rebuild my kernel.
Finally, I was able to run all the tools. After a couple of hours of running, the script determined the correct voltages for my system. Much to my disappointment, apparently the voltages for the lowest frequencies of 800 and 1600 MHz are locked on my CPU, meaning that they wouldn't go below a certain unknown level. Thus, the only advantage I can get is when the CPU is running at the highest frequency, which doesn't really happen that often, especially not in the power-saving scenarios I was hoping for.
So, the results were quite disappointing - idle temperature and power consumption remained absolutely the same, just as I thought they would. Running at the highest 2.53GHz saw a drop in power consumption from 38.8W to 36W and in temperature from 77C to 70C. Which is of course, not bad in itself, but again, not quite the change I was hoping for. Especially if you take into account the potentially decreased stability of the system which may show itself at any time.
So, the quest for power saving continues.
Finally, I was able to run all the tools. After a couple of hours of running, the script determined the correct voltages for my system. Much to my disappointment, apparently the voltages for the lowest frequencies of 800 and 1600 MHz are locked on my CPU, meaning that they wouldn't go below a certain unknown level. Thus, the only advantage I can get is when the CPU is running at the highest frequency, which doesn't really happen that often, especially not in the power-saving scenarios I was hoping for.
So, the results were quite disappointing - idle temperature and power consumption remained absolutely the same, just as I thought they would. Running at the highest 2.53GHz saw a drop in power consumption from 38.8W to 36W and in temperature from 77C to 70C. Which is of course, not bad in itself, but again, not quite the change I was hoping for. Especially if you take into account the potentially decreased stability of the system which may show itself at any time.
So, the quest for power saving continues.
No comments:
Post a Comment