[ExI] idear question

[Jason Resch]

On Sun, Nov 30, 2025, 9:59 AM spike jones via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

>
>
> Question please for microprocessor gurus please.
>
>
>
> In the olden days, we fooled around with overclocking processors, above
> their stunning 1.4 MEGAHERTZ (hey, I did say OLDEN days (a decade before we
> were giga-ing anything (when plenty of us were still kiloing about in our
> quirky electro-mathematical hobby.)))
>
>
>
> We knew what happens when we overclocked a processor: it got hotter and
> might still work up to perhaps 2 MHz if a prole was clever with his
> heatsinks and such, but it might crash or glitch and if the misguided prole
> is running Lucas Lehmers, the checksum indicates he was a bit too ambitious.
>
>
>
> We also knew back in the olden days one could go downclock: run the
> processor slower than spec.  We did it just to verify it could be done, not
> that it would have any practical purpose.  Now it might have a practical
> purpose to downclock.
>
>
>
> What I don’t know is if modern high-end GPUs can be underclocked
> reliably.  My reasoning suggests they can be underclocked.  There is no
> apparent reason why not, but please verify, for my experience came from the
> days when the clock circuitry was external (Motorola did it that way
> intentionally before the venerable 6502.)  I don’t know that it is that it
> is done that way today, that it is even possible to underclock a modern
> microprocessor.  If so, is it done the old-fashioned way: oscillator to a
> signal generator which Fourier square-wave-izes the signal and feeds back
> into the (defeated) processor clock?
>


I don't know about GPUs but modern CPUs often under clock themselves on the
fly to reduce power consumption.

See: https://en.wikipedia.org/wiki/Dynamic_frequency_scaling

>
>
> If a modern processor cannot be underclocked because there is no
> user-accessibility to the timing circuitry, but a prole has the idea(r) of
> using a microprocessor as intelligent cabin heaters in electric cars,
> replacing a resistance coil (which seems like an egregiously wasteful path
> to high entropy) perhaps there is a software workaround.  If the EV cabin
> heater is turned open throttle full heat, then the processor would work at
> maximum capacity.  If it is only a little cool outside and the prole wants
> only a tenth the heat, then the processor could do bursts of calculation:
> mine for BitCoins one second, idle for ten seconds, repeat, result low heat
> to EV cabin.
>
Heating the cabin most efficiently would still use something like a heat
pump, which can be over 100% efficient when compared to generating heat by
resistive heating (which the CPU/GPU essentially does). 300-400% efficiency
is typical. That is, for 1 kWh of energy, a heat pump can move 3-4 kWh
hours of heat energy into a given space.

https://en.wikipedia.org/wiki/Heat_pump

Jason



>
> After I wrote that paragraph above, the workaround notion grew on me: EV
> cabin heater control via software is a better idear than attempting to
> underclock a modern GPU: it doesn’t require expensive hardware modification.
>
>
>
> Sheesh, I answered my own question just by asking it.  It’s embarraskin I
> tells ya.
>
>
>
> spike
> _______________________________________________
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20251130/53476f54/attachment.htm>