[geeks] CPU fans

[Lionel Peterson]

On Oct 30, 2009, at 4:51 PM, nate at portents.com wrote:

>> I (personally) don't understand $75 heatsink/fans, but I also don't
>> overclock...
>
> $75 heatsink/fans are overkill, but a good $30 one will take a 2.4Ghz
> Q6600 (G0 rev) and allow you to run it at 3.6Ghz with just a tiny  
> bump in
> voltage.
>
> Most $75 heatsink/fans are overkill, but people buy them anyway, so  
> they
> continue to make them.
>
> BTW, if you haven't been following, the new Intel architectures  
> (Core i5,
> i7, etc.) have a sort of auto-overclock built into them.  Intel  
> improved
> the thermal monitoring, so the CPU tracks it's overall thermal
> dissipation, and as long as it's under the upper bound set for the  
> CPU, it
> will increase the clocks to individual cores when the workload isn't  
> using
> all of them (so single-threaded, dual-threaded, and triple-threaded  
> tasks
> get a speed bump on a quad-core system).  It's interesting, and not  
> a bad idea.

So they are (in effect) self-overclocking, while constantly testing  
the temp of the chip?

Interesting - I wonder how perceptible the bursts of higher speeds  
are, on a normal 'GUI' user workload?

So independently each core can run much faster, but when all four  
cores are running full out the printed speed is the best compromise  
speed...

Neat, thanks.

Lionel