[Tfug] "Downgrading" ("underclocking?") processors

[Bexley Hall]

Hi Zack,

On 2/17/2014 9:07 PM, Zack Williams wrote:
> On Mon, Feb 17, 2014 at 12:11 PM, Bexley Hall<bexley401 at yahoo.com>  wrote:
>> While lots of folks *upgrade* CPU's, does anyone have any
>> experience with *downgrading* a CPU?
>
> You might look up "undervolting" on various PC forums - the power (and
> thus heat) is the square of the voltage used, so lowering the clock
> speed and voltage results in less power usage.
>
> Most of the references I've found are almost entirely Windows based though.

<frown>  I think that's tempting fate if the design didn't *plan*
on this before hand (and, you *won't* have access to schematics
or any other documents for most "commercial" products).

Would you use this technique on an *enterprise* server?  How well
would you sleep thereafter?  :>  Would you ever know if you'd
introduced an *undetected* transient failure in the system?  (i.e.,
memory has ECC but will your CPU know that *it* just erroneously
computed 3+4=6?)

I was thinking more in terms of replacing CPU with a "lower grade"
CPU -- the opposite process of replacing the CPU with a *higher*
grade (additional cores, higher clock frequency, etc.).

E.g., the SBC's that I use for my automation system prototype are
1GHz, 1GB "PC"s.  Yet, work well with passive cooling.  Other
similarly spec'd boxes have noisey fans, etc.  The thought being
to "simply" replace CPU with pin-compatible devices and "adequate"
passive radiators to reduce the power thrown off as heat (trading
performance for BTU's).

But, it seems most folks go the other way:  trading more heat for
a faster core, etc.