Is it true that a hot CPU uses more Juice than a cool CPU?

Thanks for the effort Jokester, it obvious from your post that there is a huge difference in our levels of engineering knowledge! :o

I'm happy to take what you say as gospel as I have no way of understanding any of this! :D

I don't know the difference between a volt, an amp or indeed a watt, or what their relationship to each other is, also TDP, resistance etc

I wonder, could you put that brain of yours to work for me and do some calculations on the following. . .

Take an E6300 chip running in two scenarios

#1 running at stock 1.86GHz, 1.28vCore actual, load temp of 45°C

#2 running at 1.86GHz, 1.06vCore actual, load temp of 32°C


How much less power (watts?) would example #2 use over a 24 hour period?

thanks for the loan of the grey matter! :)

Going from 1,28 V to 1,06 V would mean a voltage drop of 17,2 %. Given it uses the same amount of Amps, it should mean you get roughly 17 % drop in power usage. However, if you use a socket power meter you will only know how much your computer uses and not the CPU alone. You also have to take your PSU's efficiency (eg. 80 %) into consideration if you want to know how much power your components draw.

Overall I only suspect your computer will draw a few percent less power in your scenario.
 
Overall I only suspect your computer will draw a few percent less power in your scenario.
Thanks sablabra,

yeah I will have to get one of those power block things and see, even though it measures the whole system load I should be able to work out how many extra watts are being used between the two difference scenarios.

If that screenshot above is accurate, I was quite suprised to realise that a stock E8400 when loaded will use almost half as much electricity as an overclock Q6600 at full load, thats like two 100w lightbulbs extra running 24/7, day in day out.

I will source me a power block thingie and check for myself, off the top of my head I will aim for a power useage of 150w when the system is idling and 200w under full load, once that is achieved I will start to work out ways to shave it down further.

Would be a cool thing to build powerful computers that can run all tasks fast and smoothly while not burning more than 100w at full load.

thanks for all the helpful replies peeps, good luck with your 5GHz overclocks! :D
 
Reading above, there's a few incorrect assumptions here, look up "temperature coefficient of resistance", in most pure metals resistance increases with temperature, but in Semiconductors the resistance actually decreases with increased temperature. Of course in power transmission for instance an increase in temperature will increase the copper wires resistance and therefore passes less current for the same supply voltage.

Unfortunately its not quite so simple in a cpu, yes a portion of the heat comes from the current passing through the copper power plain, so a lower temperature will help reduce the waste heat a bit. But this shouldn't be confused with the increased stability from lowering the silicon temperature.

Its complicated and I've not read up on it in years. But there's a few things happening in at the transistor gate level as temps increase. The ones I can remember are an increased current 'leakage' (aka static power, the power used when doing nothing i.e idle) leading to greater gain and in turn more current, a vicious circle leading to 'thermal runaway'. The other is for a given voltage the transistors switching rate reduces as temps increase. That obviously causes delays then errors in the circuits.
 
Last edited:
Hi again,

I'm trying to get one of those things that fit inbetween the computer and the mains and show me how much power is being used but I have no idea what this device is called? :o

Can anyone tell me what I should be looking for please! :D
 
Hi again,

I'm trying to get one of those things that fit inbetween the computer and the mains and show me how much power is being used but I have no idea what this device is called? :o

Can anyone tell me what I should be looking for please! :D
All these searches come up with them ;)

Electricity Meter
Power Meter
Energy Meter

Electricity Monitor
Power Monitor
Energy Monitor

I use one of these Zalman ZM-MFC2 Multi Fan Controller http://www.overclockers.co.uk/showproduct.php?prodid=BB-000-ZA&groupid=701&catid=57&subcat=189

Zalman ZM-MFC2 Multi Fan Controller

- Real-time display of overall PC power consumption
- Four sensors for temperature monitoring and display
- Monitoring and control of 3 standard fans and 1 PWM fan
- Alarm system to notify non-operation of any of the fans
Top is PC power useage (230watts)
DSC00233.jpg
 
Last edited:
Now how does that zalman thing accurately measure the computer's power consumption??

I checked into it now, and there's a little box you plug the power cord into.
 
Last edited:
Ok since theres some confusion i will explain this, an easy way and an indepth way. fornowagain is on the right track :)

easy way:
Semiconductors when heated have more thermal energy, the extra thermal energy gets the electrons more excited and so they break away from there bond which results in a free electron, as we know current is the flow of electrons through a materaial, so if giving more heat allows the meterial to have more free electrons then it will have a lower resistivity which results in a lower resistance which again allows higher currents, the higher current will produce more heat in turn however there is a limit to which this can happen, the Fermi level (look it up in wiki or sumthin) can explain this.

indepth way:
Ok here goes, semiconductors are made up of insulators and conductors(hence the semi) they are Intrinsic semiconductors (have no doping) there outer electron layer has 4(usually) electrons(take silicon or germnanium), Both silicon and germanium have valency 4, and they crystallise in the diamond structure by covalent bonding of the four valence electrons each one valence electron from its four nearest neighbours.
At room temperature some of the valence electrons acquire enough thermal energy to break away from the valence shell(away from the bond) thus giving a free electron and a hole(a vacancy for an electron in the valence shell).
This hole being a vacancy for an electron consists of a missing negative charge and is equivalent to the presence of a positiver charge. The materia, remains overall neutral since breaking each bond produces one electron and one hole, and therefore the number of conduction electrons per unit volume(ne) and the number of holes per unit volume(nh) are equal(nh=ne only for intrinsic materials))
Current can therefore be carried through the material by the motion of the free conduction lectrons in one direction and the motion of holes in the opposite direction.
The energy required to break a bond and so generate an electron hole pair is different for each material.
silicon=1.12eV germanium=0.66eV
eV is a unit of energy and 1eV= 1.6x10^-19 Joules
So from this at any temperaure (T in Kelvins) the average thermal energy of a particle is given by kT where k is Boltzman constant. if the the given temperature is high enough to allow the thermal enegy to be higher than the energy to break a bond ie 1.12eV for silicon then the increased thermal energy will break more bonds to generate more electron hole pairs and so the resistivity decreases.
Obviously a lower resistivity will create a lower resistance and therfore a higher current but there is a limit, as eventually all the covalent bonds will be brocken and all the electrons will be free so there is an optimal temperature where all the electrons will be free and then applying more heat or thermal energy wont decrease the resistivity any further, this can be shown in the Fermi function.
This is for intrinsic semiconductors however and is only the tip of the iceberg it gets a lot more complicated when using extrinsic semicondutors (p-n junctions etc) eventually after the temperature is so high, in a semiconductor junction (uses 2 types of extrinsic materials) which is what transistors and processors are made of, the junction breaks down and the current flows either way and destroys the gate. Also, higher thermal energy increases gate leakage, dusnt alow the gate to operate correctly at high frequencies beacuse of the MILLER effect which basically is due to capacitance between the gate and drain and not forgetting Thermal noise, shot noise, 1/f noise etc
Extrinsic semiconductors have other metals fused into the structures to alter the properties however i wont go into that unless u want me too lol

hope this clears things up :D (sorry for my bad typing and spelling)
 
Last edited:
Pretty sure about 70% or so of the heat from a modern CPU is a result of the current flow in the copper interconnects, there is something like 30 feet of copper "wiring" in there.
 
^^ Yeah most of the heat will be due to the interconnects, wonder if they would make cpu's with silver interconnects instead of copper, would lower the resistance quite a bit and therfore power, wouldnt like to see the price right enough hehe :D
 
Back
Top Bottom