Processor speed relative to temperature

[supasoaker]

I know it doesnst affect the actual performance as such, but doesnt high temp cause the transistors to make mistakes and cause the errors you get when the temp is too high? This might be sort of OT but Im just interested really.

seems you are bang on - read the end of the linked article - i think you may be referring to 'leakage' in the transistors that increases with temperature.

http://www.madshrimps.be/?action=getarticle&number=7&artpage=2967&articID=636

to save the hastle (and save the info - i hate it when links don't work!!):

'We can divide the total consumed power into two main parts, static power (Ps) and dynamic power (Pd).

The static power consumption is what we usually call the leakage. In an ideal transistor, it should completely shut off the channel between the source-drain, gate-source and gate-drain. Transistors are far from ideal, and the current leaks between these parts and the substrate of the processor, and this is heavily dependent on the temperature.For example, going from room temperature to 85C (~60C difference) increases the leakage power by a factor of more than 50. Thus, reducing the temperature with the same amount will make a huge impact on Ps.

Dynamic power consumption is emitted during the short amount of time that the transistor switches. Lower temperature reduces the resistance in the processor which results in shorter delay/faster switching of the transistors. Shorter delays and less noisy signals also reduce Pd.

I hope this explanation give you some clarity to the relation between power consumption and temperature. This can even be seen with air cooling: The power consumption is lower just after a load is applied compared to after a while when the temperature has levelled out, even though the load is the same.'

 

[fornowagain]

Leakage just refers to the power used when the cpu is doing nothing. As temps increase so does the leakage, it also scales with frequency, so effects overclocking there as well. Eventually you can run into a phenomenon called Thermal Runaway, extra temps increase leakage which increases temps and so on until...

Let me inverse my earlier statement. Higher temps = higher resistance. Higher resistance means slower capacitance charging and more thermal noise. Which gives lower switching rates and poorer signals. Which in turn decreases stability.

 

[Bigdogbmx]

Interesting stuff, thanks very much for the in depth explains
I think I must have read something similar a long time ago, I just couldnt remember if I had it right. I do start to flake out a bit when people use algebra though, its a long time since I did any maths other than say, I have 200 pounds and my bus ticket is 70 pounds, how much crap can I buy with remaining 130 pounds.

 

[death_looms]

No equation, but electricity does travel faster at cooler temperatures.

Emm, no.

What actually happens when you apply a potential across say a wire, is that an electric field is set up inside the wire (actually along the edges - skin effect). This may be thought of as the 'speed' of electricity, which will be the speed of light in that wire. The electrons move because they are subject to a force in the electric field.

When the wire heats up, the atoms in it will vibrate more, causing more electrons to impact them, transferring energy (results in heat). Thus the resistance is increased. The 'speed' of the electricity is the same, the CURRENT is reduced.

Clarified any?

 

[JonJ678]

^I thought he meant drift velocity, not current

p.s. I didn't abandon the last thread I saw you in as such, just completely ran out of things to say and didn't want to conclude a useful thread with "thanks" in case someone else added to it

 

[death_looms]

Depends on the definition of 'electricity speed', but I totally see where you're coming from! As for the last thread, I hope I managed to sort out any confusion. Seeing as I plan to teach physics in 20 or so years time, I better keep trying to explain myself more clearly than I do at the moment!

After a quick scan of the thread again I see power consumption mentioned. Power consumption is proportional to current (squared) and the resistance. As resistance increases with temperature, thus the power consumption will increase linearly with it Temperature won't affect the speed (unless it makes the core unstable, meaning instructions must be run several times until a result is error free), but it will affect power consumption. Keep 'er cool lads

 

[Monkeynut]

Emm, no.

What actually happens when you apply a potential across say a wire, is that an electric field is set up inside the wire (actually along the edges - skin effect). This may be thought of as the 'speed' of electricity, which will be the speed of light in that wire. The electrons move because they are subject to a force in the electric field.

When the wire heats up, the atoms in it will vibrate more, causing more electrons to impact them, transferring energy (results in heat). Thus the resistance is increased. The 'speed' of the electricity is the same, the CURRENT is reduced.

Clarified any?

So to get the same amount of current out of the other end you need to put more current in?

 

[JonJ678]

Nah. Charge is conserved.

If the potential (=eletric field in the wire) is the same, then the force driving current down the wire is the same. However the resistance to the flow increases with temperature, so the same force gets less current to flow.

Current is very like water in a loop. if the tubing diameter decreases, say you clamp it, the flow rate goes down. But you aren't losing water anywhere, and the pump is working just as hard. So if the resistance to current goes up, the charge will flow more slowly. Current is rate of flow of charge, like flow rate is rate of flow of water.

I dont quite see the analogy between electric field and velocity that death is drawing here, but electrons definitely aren't vanishing partway down the wire.


I always assumed that the transitors in a processor were wired together using semiconductor of some sort, however recently saw an intel presentation suggesting copper is used instead. Anyone know which it is?

 

[death_looms]

Fairly accurate there

The electric field is what makes the electrons move, the potential difference sets up the electric field. Creating an electric field is not instantaneous, rather the field will propogate (move) along the wire. It can only do this at the speed of light, which can be seen as immediate for most applications on earth! So if you have a MASSIVE circuit with two light bulbs in it and you switch on the current, one will light up before the other, as the electric field will reach one first (which causes the electrons to move).

In real-world this makes little to no difference as the limiting factor is resistance losses, which inhibit the building of such large circuits.

better?

@Monkeynut: If you're loosing current due to resistance increasing, the only way to increase it is to increase the potential driving the electrons (which will heat it up more!)