Processor speed relative to temperature

[supasoaker]

i'm sure it will vary from chip to chip but does anyone know how temperature affects the electrical resistivity of the chip?

anyone got an equation for this??

I'm trying to calculate how much faster or slower my CPU would get depending on how hot or cold it is..........

big thanks to anyone that can help!!!!

 

[supasoaker]

No equation, but electricity does travel faster at cooler temperatures. How this translates to actual performance however, i'm not sure.

yeah, thats exactly what i'm looking for, was hoping someone working closer to the industry or sommin might be on this forum and know.........

i've seen videos where people take the heatsink off and wait for it to freeze........it has to be a measured amount, i just dunno where to find it, been looking but still no luck..............

 

[supasoaker]

Based on?

Cpu speed doesnt work like that. It goes through say 3000 million clock cycles a second, regardless of temperature. It calculates a certain amount on each clock cycle.

The reason cold processors are quicker is simply that they can be run faster without becoming unstable. The cold itself in no way increases speed.

Taking the heatsink off and waiting for the processor to freeze is ill advised

it does if you super freeze it then the components become superconductors.

i see your point and think you are correct, but it still is a measurable amount relative to the temperature...........there may be bands where it has more or less effect, electron flow is still relative to temperature though..........

if you wanna know why i'm asking i'm currently designing a cooling system and need to look into what bands of temperature are most efficent to operate the chip at.................as a general rule ambient temperature is best i find as this is the coolest I might be able to efficently cool the CPU down to, below ambient would consume too much power...............also lower wear and tear as if i keep the chip at ambient temperature (i wish) there will be less expansion and contraction as it warms and cools therefore (hopefully) lowering wear.........................won't help against Black's Equation though!!! (might as there will be less heat energy to affect the magnetic fields ......... :S ............)

 

[supasoaker]

This is AC. The electrons aren't going anywhere much. Even if they were, drift velocity is hardly swift or related to processing speed. This effect does not exist unless I'm very much mistaken and Superewza actually manages to back up his post. Consequently no, it is certainly not measurable.

If you cool a processor enough it does not become superconducting. It's an impure piece of silicon, there's not a hope it'll superconduct. Even if it somehow did so, I challange you to think of a way take a 200W heater down to -270K

a/c gets changed into dc though...............aren't they going somewhere then?????

its also bad to cool it that much as it can become brittle.............i'd give it a go, why not ey.................men have been trying for years who says i carn't............I know of someone who managed to maybe generate enough cold and that was to watch light move slowly............

these questions are good on the forum though look at all the info that comes out................how many ppl in this post work in IT or electronics???? hehe......

 

[supasoaker]

I don't think there's very much DC in a processor, suspect none beyond the power lines. Where magically faster electrons aren't going to help.

Doesnt matter how brittle it is, silicon is brittle anyway. Semiconductors are very different to superconductors. The problems are that your initial premise is flawed and somewhat nonsensical and that physics doesn't work the way Superewza thinks it does,

For the purpose of designing a cooling system, your OP still makes no sense I fear. Now, if you want to overclock it further, a better cooling system would be excellent. 920s run at about 5ghz on phase cooling, which is rather quicker than stock. Care to share the intial ideas to the design? I think you'll do better aiming for 'as cold as possible' than trying to get it loading at -30 to -25 degrees.

well believe it or not its not for overclocking so i'm not looking for massive performance..........i'll be running stock initially on a thermpsiphon watercooling so its silent, seen it done before and can be made to be quite good..........i got a mini fridge i can move next to my pc later which consumes 100w every so often (love the pulsing!!!) so might move on with that but i'll see - main thing is to get the thermosiphon running well for now.

ones i have seen can keep 40 deg. C constant which is quite good really, its silent and increases the thermal efficiency of my pc (small savings but a fun project.........its obvious my friday nights are well spent!!! ).

 

[supasoaker]

You cannot cool a pc using a mini fridge. Or a big one for that matter. Dropping ice cubes in the reservoir can be good, but all you'll manage here is a dead fridge.

My point is that the only increase in speed comes from overclocking. If staying at stock all you'll achieve is a lot of time spent and some entertainment. If you want a silent cpu cooler just stick a passive heatsink on it.Your computer wont be silent unless you manage to do something about all the fans, not just the processor one.

This all sounds quite negative. I'm not actively trying to be discouraging, it's that I dont see what you're trying to achieve and want to be clear that your computer is not going to be any quicker just because its cold

lol at Superewza completely failing to back up his claim

haha now i see, i wasn't saying that the chip will run faster if it is cold, you are dead right bout the clock cycles and stuff.............even though if cold it may run more efficiently (sorry but had to add that!!).

yep, the point is to remove the fans and have it silent, last one i built has lots of fans and is too much for me even though its nice and cool.

the fridge is way out there in my plans, i'd only use it to aid the cooling not rely on it cos as you say, it could easily break it.

ever heard of (i'm sure you have) the vapochill units? expensive but similar to a fridge!!!

this whole thing came about as i am trying to learn at the same time and have a better understanding of the dynamics of the system i'm building as well as cooling in general....................and in all this somewhere is that it would possibly be more efficent for me to tune my cooling system to certain temperatures.........luckily for me 40 odd degrees Celcius seems a good enough target for the plug and play items pumped out of the factory, carn't hurt to ask though.........

 

[supasoaker]

If cold it will indeed run marginally more efficiently, though the savings from this will have to be offset against the cost of the cooling and it won't balance

Before putting much effort into the cpu fan, try to work out what to do with the psu one. I periodically purge the fans from my computer and always end up with the psu one present and annoying. 400W fanless psu or the epic koolance ones seem to be the only standard options.

Vapochill is to fridge as ducati is to bicycle. Phase cooling is refrigeration done 'right' as far as computers are concerned. The difference is the wattage it can dissipate, a fridge is aimed to slowly cool the contents then rely on insulation to hold it there. If you put a heater in a fridge it dies quite swiftly, the compressor can't run flat out all the time without failing. If you pump water through the fridge, say with a radiator in it, I'd be curious as to how long it lasts.

I recommend water cooling and overclocking. The former has forced me to learn considerable amounts of thermodynamics/fluid dynamics and the latter an aweful lot about computer hardware. Its good learning experience, though making your own cooling system would also be excellent it may be a little much to start off on. Do a build log if you do choose to make it yourself please

i'm with you on the temperature, i'm also tyring to make it energy effiecent which is why i have set ambient to be the lowest i would want to achieve.

i currently have a zalman which i love to bits and does the job for me. i'd put one in a box and run water through it if i could be arsed but thats a bit much. got a OCZ Hydroflow for my current project.........reasonably priced and i liked the design for thermosiphon cooling.

i'm also incorporating hydraulics into the design to try and increase the rate of flow, got an automotive heater matrix as my radiator, just waiting for some other parts to arrive before i assemble which should be in a month or so...........tyring to do some math behind it all but it can take its time with other pressures in life............lol

i'll pop it up on the forum when its all together no prob, would be good to get some feedback..........all insults welcome too!!! hehe...........

 

[supasoaker]

In the copper metal traces the resistance increases with temperature. For the doped semiconductors its more complicated. Having odd temperature profiles, all depending on electron mobility through carriers. From 0K upwards they decrease sharply in resistance. A bit further on as temps increase the resistance increases like a metal. At high temps they behave like intrinsic semiconductors where resistance decreases with temperature. So depends on the doping and temp, but in the bit your looking at the resistivity is inversely proportional to the electron mobility where temperature is a function.


Resistance vs temp with increasing doping levels.

To simplify lower temps = lower resistance. Lower resistance means better capacitance charging and less thermal noise. Which gives greater switching rates and cleaner signals. Which in turn increases stability.

If you want the equations read on, don't ask me to explain it though > http://www.cactusconnects.co.uk/Investigation_into_low_temperature_resistance.htm

It's DC, the cpu works from what is known as a power plane, usually labeled Vcc. But its a + voltage wrt to ground.

thanks - great article!!! - exactly the type of thing i was looking for.

from what i managed to gather it mostly tests copper, constantan and germanium - are these the primary components of a CPU in terms of metal i thought tin (as a filler) would be quite big up there and of course silicon...............

i know copper, constantan and germanium are big in the industry but any ideas on how much would go into a chip???

 

[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.'