AMD Cool'n'Quiet (PowerNow!) for Windows XP (SP2) and Linux (Fedora 3)

An Unscientific Experiment...

I decided to try to acquire more meaningful data. My guess was that the CPU temperature should stabilize after an hour of idling at the same settings. I made sure that the screen saver was turned off, and that the power settings would not shut off the disks, monitor, etc. for the duration of the test.

The only variable in the experiment was the power management scheme. I didn't add a test for temperature while the chip was under stress for the simple reason that I didn't care about the results. If I'm playing a game, I want the CPU to use as much power as it needs. Who cares if that's 89W or 69W.

So, I started the experiment at 10:07 PM with the following values:

CPU Temperature: 34 Degrees C.
Sys Temperature: 33 Degrees C.
CPU Fan Speed: 3125 RPM

These values represent what happened to be on the display after playing with the power scheme for a few minutes, so they shouldn't be associated with either scheme.

I set the power scheme to 'Home/Office Desk', watched the CPU temperature rise to about 35 degress, then went upstairs to watch TV. When I came back to check the readings at 11:09, I fould this:

CPU Temperature: 37 Degrees C.
Sys Temperature: 35 Degrees C.
CPU Fan Speed: 3125 RPM

At 11:10 PM, I turned on Cool'n'Quiet, and walked out of the room to watch TV for about another hour. When I returned at 12:09, I found these readings:

CPU Temperature: 33 Degrees C.
Sys Temperature: 35 Degrees C.
CPU Fan Speed: 3125 RPM

You could certainly complain that the sample set is too small, and that there are some aspects of the data that should be explained, but I wouldn't listen. After all, I'm not gunning for the Nobel Prize with this. I'm just offering supporting data for my conclusions.

... With Unscientific Conclusions

The most important aspect of this feature is in something that the numbers don't tell you. Once you set up the driver and the power scheme, the power savings cost almost nothing in performance.

In my experience, I wouldn't have known that I was in a power saving mode if I hadn't been watching the performance meter and the Core Center readouts. Though I can think of lots of scenarios where the power saving will be minimal, I can't think of any reason why Cool'n'Quiet should be avoided. In the time that I have had it running, I have not found a reason to turn it off.

Someday, maybe I'll find the time and energy to do a proper experiment. Maybe. Someday.

<< Page 2 | Page 4: Onward to Linux >>