If concerned about CPU temperature or reducing electricity use, it's a simple thing to "Under-Clock" your machine... When the machine boots, you can typically enter the bios at that time and change the clocking. It's pretty easy. If interested in changing the CPU clock speed, GOOGLE a search for your Motherboard. Example: " asus motherboard change clock speed ". By under-clocking my machine, I reduced its temperature by 11 degrees and see no difference in performance. If ever you find a need for speed, just change it back. As for the project, I haven't observed much difference in the stats (Not that it matters much to me).

There are two variations of this that can be used by themselves or in tandem:

The amount of heat/electricity that your computer produces/consumes is approximately = MHz * Voltage^2

By changing the clock multiplier on your system you can reduce the speed (Mhz). Going from 2Ghz to 1Ghz should cut your CPU power usage in half (keep in mind it won't change the power the rest of your computer uses, so your power bill won't be cut in half). This option reduces the speed of your computer, so if you cut your speed in half you will cut your credits/day in half.

The other option is to change the voltage. CPU makers have to set the voltage of a given processor line at a value each chip will run at. Most of the chips can run with less voltage. For example, if a given chip runs at 1.35 volts, it may be able to run at 1.25 volts. If stable at this speed, your machine will be just as fast but will be cooler and use less electricity (in this example about 15% less).

Some folks use both options to allow their computers to run so cool they can run off passive heatsinks (no fans)!

Overclocking is the exact opposite of these, using higher voltages and multipliers/fsb speeds to get more performance out of a CPU at the cost of more heat and electricity.

---

There are two variations of this that can be used by themselves or in tandem:

The amount of heat/electricity that your computer produces/consumes is approximately = MHz * Voltage^2

By changing the clock multiplier on your system you can reduce the speed (Mhz). Going from 2Ghz to 1Ghz should cut your CPU power usage in half (keep in mind it won't change the power the rest of your computer uses, so your power bill won't be cut in half). ...

There is little point in just cutting the speed. Once the speed is reduced the voltage can also be reduced. That way you get a reduction in power that is more significant than the reduction in speed.

One way to think of this is to notice that when a circuit component changes state it takes a certain (large) number of electrons to do this. To get those electrons in quick you need to push them harder; if you allow more time you don't need to push so hard.

River~~

---

There is little point in just cutting the speed. Once the speed is reduced the voltage can also be reduced. That way you get a reduction in power that is more significant than the reduction in speed.

But the best way is to just cut the voltage ;) I dropped my dothan 1.86GHz laptop CPU by 1/3 and it dropped the temps and increased the battery life massively.

RM clock is great - there are a few other tools too.

---

I use Notebook Hardware Control (NHC - use to be called CHC). http://www.pbus-167.com/
Worked great on my Pentium M! Voltages were dropped by quite a bit. Basically I found the lowest voltage that worked at max CPU speed that was stable tested at max CPU use, then added a bit extra voltage "just in case". Now at max CPU use, it is quieter and cooler then when it was only working at 40% cpu use. (by the way, don't have it set the new voltage on windows startup until you are really sure it is stable!)

There is little point in just cutting the speed. Once the speed is reduced the voltage can also be reduced. That way you get a reduction in power that is more significant than the reduction in speed.

But the best way is to just cut the voltage ;) I dropped my dothan 1.86GHz laptop CPU by 1/3 and it dropped the temps and increased the battery life massively.

RM clock is great - there are a few other tools too.

Cutting voltage without cutting speed reduces the tolerance - ie increases the chances of a small number of miscalculations.

As always it is a trade off - this time a three way trade off between reliability, throughput, and energy used per flop.

That is why my choice would be to cut volts and speed together -- that keeps the reliablity up and reduces power by a greater factor than the speed (so that the energy used per flop drops).

But it is like eating yoghurt after the sell by date - there is no exact point where you can ay it becomes unreliable, and the final judgment is personal. Tarx does about what I would do - find where it just works, then add back a tad for safety.

R~~

There is little point in just cutting the speed. Once the speed is reduced the voltage can also be reduced. That way you get a reduction in power that is more significant than the reduction in speed.

But the best way is to just cut the voltage ;) I dropped my dothan 1.86GHz laptop CPU by 1/3 and it dropped the temps and increased the battery life massively.

RM clock is great - there are a few other tools too.

Cutting voltage without cutting speed reduces the tolerance - ie increases the chances of a small number of miscalculations.

As always it is a trade off - this time a three way trade off between reliability, throughput, and energy used per flop.

That is why my choice would be to cut volts and speed together -- that keeps the reliablity up and reduces power by a greater factor than the speed (so that the energy used per flop drops).

But it is like eating yoghurt after the sell by date - there is no exact point where you can ay it becomes unreliable, and the final judgment is personal. Tarx does about what I would do - find where it just works, then add back a tad for safety.

R~~

Not to mention that individual CPU's (like yoghurt pots) will be slightly different, some will work better at a lower voltage than others (or not go bad after the sell-by-date as quickly as others, in the case of yoghurt). Even within one single wafer produced on the same day, not all chips are exactly equal. Some will be better, others will be not so good. Finally, sometimes companies "down-bin" chips, which means tha something that actually pass a higher spec is sold at a lower spec - because that's what was requested by the customer(s).

Lowering the voltage will help more than lowering the frequency, as the oversimplified calculation for CPU power consumption is P = C * f * v * v, where C is some constant for the device, f is the frequency, and v is the voltage. So take 10% of the voltage and you get a 19% less power (0.9 * 0.9 = 0.81 = 19% better), whilst dropping the frequency on it's own by 10% only lowers the power by 10%. Of course, doing all three at the same time will have an even better effect on the power consumption.

As River stated, lowering both is the safer option.

--
Mats

---

...
But it is like eating yoghurt after the sell by date - there is no exact point where you can ay it becomes unreliable, ...

Not to mention that individual CPU's (like yoghurt pots) will be slightly different, some will work better at a lower voltage than others (or not go bad after the sell-by-date as quickly as others, in the case of yoghurt).

and every responsible manufacturer (of cpus or yoghurts) adds in a safety margin, and if you then exceed their recommendation the first thing you do is erode that safety factor, putting you in the zone where it is probsbly safe but you can't complain if it is not. And of course you can't sue.

Finally, sometimes companies "down-bin" chips, which means tha something that actually pass a higher spec is sold at a lower spec - because that's what was requested by the customer(s).

or something that only just misses a higher spec is sold at the next lower - I got a 100MHz cpu once (486-DX iirr) that would run at 183 but not at 197. I reckon it was made as a 200MHz and failed quality control, but Intel figured they'd like to make something on it rather than put it in the bin.
R~~