clock speed vs. heat produced

afrothunder · Sep 5, 2007

I dont know if this is supposed to be common knowledge, but how does the heat produced (or power dissipated) from a CPU increase with clock speed?

for example, if a CPU at 2ghz produced 100W of heat, an i overclocked it to 3ghz (while keeping the voltage and the temperature constant), would it produce 150W?

i've read before that heat production is linear with respect to core voltage, but maybe i'm not sure about that either.

thanks.

th3 · Sep 5, 2007

afrothunder said:
for example, if a CPU at 2ghz produced 100W of heat, an i overclocked it to 3ghz (while keeping the voltage and the temperature constant), would it produce 150W?

Each transistor will consume a certain amount of power when switching on/off, the more often they switch the more power they will consume. Switching power is just one of the factors of a CPUs total power consumption, higher clockspeed alone will not account for a huge increase in total power draw. (i have no idea how much it would be, i guess an electrical engineer could come up with estimated numbers)

However, when the transistors get overclocked to the point that they cant keep up with the speed, and you need to increase vcore, then the power used for switching is increasing also per switch operation, that combined with 50% more switch operations as in your example will give a more significant increase in power draw.

aja · Sep 6, 2007

afrothunder said:
I dont know if this is supposed to be common knowledge, but how does the heat produced (or power dissipated) from a CPU increase with clock speed?

for example, if a CPU at 2ghz produced 100W of heat, an i overclocked it to 3ghz (while keeping the voltage and the temperature constant), would it produce 150W?

i've read before that heat production is linear with respect to core voltage, but maybe i'm not sure about that either.

thanks.

Power=voltage x current = an indication of heat produced, depending on efficiency. Therefore the heat produced is linearly proportional to the voltage.

But at the same time it depends on the load the CPU is placed under(hence the low idle temperatures) as well as the clockspeed, as th3 explained

DualSnakeyes · Sep 7, 2007

But bumping a cpu from 1.3V to 1.4V does take more Power (More than you think)

I do not have the link but i know i did a PSU Requirement Calculator and it asked me for my CPU and if its OCed so on default it was in the 80's->90's i think and at my PC it was 110W
and im only pushing it to 1.35Vcore and 3.2GHz

One Bull · Sep 7, 2007

Well it's not totally linear from my experience.
I noticed as the speed increased while keeping the voltage the same there is more heat produced. But it's not 50% more with a 50% oc

bing · Sep 7, 2007

Voltage -> Temperature is not linear, but ^2.

DualSnakeyes · Sep 7, 2007

Temps are also determined by your Room Temperature if its 33C your CPu aint gona be running at 40C at a 50% OC

If you have Water cooling its the same thing the heat has to be removed from the water and if there is only hot air your temps are gona be high.

I run in the 40's under heavy load (Prime95) think it peaks at 43C and thats in Temps Ranging from 20C to 28C So even if i have a slight Over clock my vcore is still 1.35 and my temps are as stated. if i ran it at 1.43 core and ran at 3.5Ghz it jumps to 48C max this on stock heatsink

aaa · Sep 7, 2007

Heat increases linearly with clockspeed.
Heat increases even faster (^2?) with voltage.

Whether the temperature also increases is heavily dependent on the heatsink you use. A more accurate way would be to measure electricity consumed.

Kuroimaho · Sep 7, 2007

bing said:
Voltage -> Temperature is not linear, but ^2.

Indeed.

Compared to that the increase from the higher freq is negligible.

Dragonprince · Sep 7, 2007

afrothunder said:
I dont know if this is supposed to be common knowledge, but how does the heat produced (or power dissipated) from a CPU increase with clock speed?

Yes.... but heat climbs much faster than speed after a certain point. Each chip is different but for current processors the C2D and Quads seem to scale significantly before heat is an issue. Recent A64 X2 chips have pushed that platform to its life cycle end where the chips do not have much left to give without significant cooling coming into play.

aaa · Sep 7, 2007

Dragonprince said:
Yes.... but heat climbs much faster than speed after a certain point. Each chip is different but for current processors the C2D and Quads seem to scale significantly before heat is an issue. Recent A64 X2 chips have pushed that platform to its life cycle end where the chips do not have much left to give without significant cooling coming into play.

Is it really speed? Those 125w Athlons are 1.35-1.4v compared to their 89w 1.3v-1.35v brethren. You have the C2Ds going far without needing a voltage change, but when you come around to doing so the heat goes up.

Dan0512 · Sep 8, 2007

A voltage increase causes an exponential increase in temperatures. When I increase my voltage from 1.35 to 1.45 it adds more than 10 degrees to the load temperature.

dab

aja · Sep 8, 2007

bing said:
Voltage -> Temperature is not linear, but ^2.

Why is this?

I am not saying that you are a liar, I just want to know why I am wrong?

aja · Sep 8, 2007

Dan0512 said:
A voltage increase causes an exponential increase in temperatures. When I increase my voltage from 1.35 to 1.45 it adds more than 10 degrees to the load temperature.

dab

Okay that is temperature, which depends highly on your cooling setup's ability to dissapate heat as well as its threshold - but what is the relationship in terms of HEAT, as in how is voltage proportional to Joules of heat produced?

bing · Sep 8, 2007

The heat is in Watt of power dissipation.

From there to temperature result depends on the thermal resistance of the hsf which is Celcius/Watt.

Edit :

Added a formula that is good enough to "estimate" the temp increase/decrease for a cpu :

Power Dissipation = PD in Watt
Voltage = Volt
Freq = Hz

Total PD in Watt = (Stock PD in Watt) X ( (CPU Voltage ^ 2) / (Stock Voltage ^ 2) ) X ( (OC Freq) / (Stock Freq) )

Cpu temperature = (Total PD in Watt) x (HSF's Thermal Resistance in C/W) + (Ambient Temp in Celcius)

Just a reminder, these are just "on paper" calculation, other factors affect the final outcome big time such as properly mounted HSF, TIM applied properly, IHS and HSF base flatness and etc.

th3 · Sep 8, 2007

aja said:
Why is this?

I am not saying that you are a liar, I just want to know why I am wrong?

To double the power consumption of a C2D with vid of 1.325, clockspeed staying the same, how much vcore do you need? Probably less than 1.6v. A linear increase would mean you needed 2.65v.

savageseb · Sep 8, 2007

bing said:
The heat is in Watt of power dissipation.

From there to temperature result depends on the thermal resistance of the hsf which is Celcius/Watt.

Edit :

Added a formula that is good enough to "estimate" the temp increase/decrease for a cpu :

Power Dissipation = PD in Watt
Voltage = Volt
Freq = Hz

Total PD in Watt = (Stock PD in Watt) X ( (CPU Voltage ^ 2) / (Stock Voltage ^ 2) ) X ( (OC Freq) / (Stock Freq) )

Cpu temperature = (Total PD in Watt) x (HSF's Thermal Resistance in C/W) + (Ambient Temp in Celcius)

Just a reminder, these are just "on paper" calculation, other factors affect the final outcome big time such as properly mounted HSF, TIM applied properly, IHS and HSF base flatness and etc.

you guys forget pressure... and air flow(speed)...

Dan0512 · Sep 8, 2007

th3 said:
To double the power consumption of a C2D with vid of 1.325, clockspeed staying the same, how much vcore do you need? Probably less than 1.6v. A linear increase would mean you needed 2.65v.

Nope, somewhere around 1.9v.

dan

th3 · Sep 8, 2007

Dan0512 said:
Nope, somewhere around 1.9v.

dan

Around 1.875v for a static load, so you may be right. Semiconductors are not that static, and i thought total resistance/impedance went down at increased voltages, so i estimated ~1.6v, cold be way too low though and i could be way wrong.

clock speed vs. heat produced

afrothunder

New Member

th3

Member

aja

Member

DualSnakeyes

Registered

One Bull

Member

bing

Low Profile Senior

DualSnakeyes

Registered

aaa

Member

Kuroimaho

Member

Dragonprince

Member

aaa

Member

Dan0512

Member

aja

Member

aja

Member

bing

Low Profile Senior

th3

Member

savageseb

Member

Dan0512

Member

th3

Member

Similar threads