[mythtv-users] Lowest power HD frontend?

[Raymond Wagner]

On 5/29/2013 9:48 AM, Jeff Siddall wrote:
> On 05/28/2013 03:46 PM, Raymond Wagner wrote:
>> Even still, you're never going to end up with an expensive pile of
>> melted plastic.  All modern x86 CPUs (a term to which Atoms do not
>> qualify) have a "boost" mode, where thermal sensors will determine the
>> amount of headroom a CPU has, and dynamically exceed its rated speed.
>> While the lower end Pentiums and Celerons have this disabled, they still
>> have the reverse where the CPU will drop below its rated speed if it has
>> exceeded its thermal limts.  You can take the heatsink off a modern CPU
>> and all it will do is downclock itself to maintain that stability 
>> margin.
>
> Hmmm... hadn't thought of that.  So you are saying I could effectively 
> throttle my system power consumption by simply removing the fan from 
> my CPU heat sink?

Absolutely not.  I'm saying your CPU won't eat itself.  There's not 
going to be anything effective about your CPU operating in this manner, 
other than it will retain some minimum level of functionality to keep 
the system running.

> That would be fantastic -- silent and low power -- but I am skeptical 
> about the longevity and reliability of a system like that.

Silent, low power, and absolutely awful performance.  This is not 
something that is a good idea.  This is merely a protection mechanism 
for the CPU should you fail to give it sufficient cooling.  Note that 
your CPU will be doing this to some degree anyway.  All removing the 
heatsink does is mean it cannot speed back up for normal operation once 
you load it up.  You're stuck down at that extremely low level of 
performance.

As for reliability, CPUs themselves are good for several hundred 
degrees, and you have to worry about the heat destroying the organic 
package long before anything might harm the silicon die.  The ultimate 
failure mechanism for most solid state electronics is electromigration.  
The metallic components slowly break down until they eventually fail and 
break the circuit, rendering the chip inoperable.  In a normal CPU under 
normal conditions, this is generally designed to be at least 10-15 years 
out, long past the system's useful life span.  Higher temperatures 
accelerate this process, but then the reduced current from the reduced 
core voltage slows it, so if anything, a CPU operating under this 
condition would have greater longevity than one operating as designed.

>  Has anyone done this before?

TomsHardware did this to a P4 almost 12 years ago to showcase the new 
ability.  The system downclocked from 2GHz to 200MHz, and Halflife 
slowed to a crawl, but otherwise kept running.  The page seems to have 
since been taken down.  You can find the video they made of it on 
Youtube, but this is the closest I could find to the article itself.

http://www.geek.com/chips/toms-hardware-burns-up-amd-546917/