[mythtv-users] silent and low-power: AMD or Intel?

[Brian Wood]

On Wednesday 27 May 2009 19:03:35 Johnny wrote:
> > Please forgive me, I havent read the past discussion.  However, this
> > makes no sense.
> >
> > If reducing the frequency reduces the CPU heat output, but does not
> > reduce the overall power consumption, where did the energy that was
> > being spent as heat go?
>
> Generally the clock frequency doesn't have the affect on power
> consumption that people assume it does (eg 1/2 clock frequency does
> not equal 1/2 the power). The power is much more dependent on the cpu
> voltage. Generally to get a CPU to run stably at a higher frequency it
> requires a higher voltage. Likewise if you underclock a CPU you can
> get it to run stably with a lower voltage. So it is the change in the
> core voltage that is primarily responsible for the change in power.
> Also it would probably be more enlightening if we had numbers while
> the cpu is fully loaded with something like prime95 or cpuburn, and
> numbers for when it is idle. I think this would tell a slightly
> different story. At idle the CPU is doing minimal work and much of the
> chip will be minimally used either way. Therefore it will be less
> affected by clock frequency. But under full load I would think the
> higher clock frequency would come into play. I am not sure about all
> of this, I am sure chips are way more complex than what I learned
> about in my undergrad systems architecture class many years ago.

What you say makes sense. The transistors on the CPU draw current when they 
change state. If the CPU is lightly loaded, not many of the devices are 
changing, so clocking down wouldn't make that much difference. If the CPU is 
being pushed hard, devices are switching as much as they can, so a clock 
reduction will have a much more noticeable effect.

Then you have to factor in the voltage changes, which effects how much evergy 
each transistor requires to change state.

Things start to get very complicated at that point, and empirical data 
(measurements) are probably the best way to see the overall situation.

I still think my Kill-a-Watt was one of the better $20 investments I have 
made.


-- 
beww
beww at beww.org