Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

CPU heat

Status
Not open for further replies.

Dr.EM

New Member
Hi. I was wondering about this, thought mabye some of you might understand it better. Basically, I wondred why it is that a CPU should get hot. As far as I knew, it dealt with 0 and 1 logic signals, so why should it get hot. If the transistors are either on or off, the power consumtion should be low? And the signals, they arn't really powering anything are they, they are just signals. Is it just that there are so many little transistors that the combined heat given off is very high? I understand why power amplifiers get hot, because they send large currents out, and the in most types, the output transistors are not saturated, so they need to dissapate the heat generated by thier resistance. Anyone understand this better :?
 
Dr.EM said:
Hi. I was wondering about this, thought mabye some of you might understand it better. Basically, I wondred why it is that a CPU should get hot. As far as I knew, it dealt with 0 and 1 logic signals, so why should it get hot. If the transistors are either on or off, the power consumtion should be low? And the signals, they arn't really powering anything are they, they are just signals. Is it just that there are so many little transistors that the combined heat given off is very high? I understand why power amplifiers get hot, because they send large currents out, and the in most types, the output transistors are not saturated, so they need to dissapate the heat generated by thier resistance. Anyone understand this better :?

It's pretty simple!.

As you say, when transistors are fully ON or OFF, they dissipate very little power, but the switching from ON to OFF, and OFF to ON, isn't instantaneous, so during that short period of time they dissipate heat.

This is why switchmode transistors usually require heatsinks!.

In the case of a CPU you have so many transistors, certainly 100's of thousands, if not millions?, so that the amount of heat dissipated becomes appreciable.

Assuming the losses during switching remain constant?, the faster a device runs the higher the percentage of time it will spend dissipating power, which is why the faster you run it, the hotter it gets.
 
Dr.EM said:
I wondred why it is that a CPU should get hot.

Just one more thing : considering PC's CPUs, in order to work (switch) faster, they must get smaller*. The number of transistors and their switching frequency increases, the surface decreases, therefore the heat density rises. So it should not be suprising, that modern CPU can have larger dissipated power density than the iron !


* - But in general, all dependencies are much more complex.
As long as CPU's performance depends on their switching frequency, one must reduce their size to avoid occuring of so called 'transmision-line phenomena'. However, it's not always like that - if you want to know more, You need to read something about semiconductors and computer architecture. Just to give some inspiration - there use to be an architecture called GAPP, which could solve certain problems faster than any other, while switching only with frequency around tens of MHz. And most circuits now works according to so called synchronic paradigm, and there's lot of interest in asynchronic ICs, which - while difficult to design - would bring heat related benefits.

I hope this gave some food for thought - I wrote it to be somehow useful, as so long I only ask questions here :?
 
It's Impossible to build transistors without building in little capacitances. When the transistors switch they alternately charge and discharge these little capacitors. This creates a net current flow from supply to ground. Take a look at the diagram of a MOSFET and you'll see why it has capacitances - the gate and bulk make a big capacitor.

One of the reasons that smaller transistors run faster it that these small capacitances shrink based on transistor length squared. A transistor's ability to drive current shrinks based on length. So as you shrink transitors their speed goes up but power dissipation goes down.

This decrease in power is one of the big reasons Intel is always trying to push the geometries smaller. The problem now is that as you make transistors smaller they leak more. At smaller than 90nm the leaking starts to dissipate more power than the swithching.
 
Antey said:
Just one more thing : considering PC's CPUs, in order to work (switch) faster, they must get smaller*. The number of transistors and their switching frequency increases, the surface decreases, therefore the heat density rises. So it should not be suprising, that modern CPU can have larger dissipated power density than the iron !
* - But in general, all dependencies are much more complex.
As long as CPU's performance depends on their switching frequency, one must reduce their size to avoid occuring of so called 'transmision-line phenomena'. However, it's not always like that - if you want to know more, You need to read something about semiconductors and computer architecture. Just to give some inspiration - there use to be an architecture called GAPP, which could solve certain problems faster than any other, while switching only with frequency around tens of MHz. And most circuits now works according to so called synchronic paradigm, and there's lot of interest in asynchronic ICs, which - while difficult to design - would bring heat related benefits.

I hope this gave some food for thought - I wrote it to be somehow useful, as so long I only ask questions here :?

Dont forget about the CPU's oscillator. It can have a large effect on speed.
 
The CPU Heat....the transistor explaination....

from
https://www.electro-tech-online.com/threads/improving-freq-counter-input-sensitivity.15672/


As you say, when transistors are fully ON or OFF, they dissipate very little power, but the switching from ON to OFF, and OFF to ON, isn't instantaneous, so during that short period of time they dissipate heat.

Does this mean that transistors dessipate p[ower only when they are in transit from one pont on the load line to another and do not dessipate heat when fixed at one point on the load line ??? :?: :?: :?:

I fail to figure out any explaination for this... If the transistor is on, it ought to let current pass through it. And it has a resistance and shall definitely get hot...

Am I wrong ?


EDIT : Wanted to make a new topic.... Got bungled up!!! :( :x
 
CPU's use CMOS (complemetary MOS) to reduce the on-state current
When in a "1" or a "0" all that is being drawn is a leakage current.

This does produce power loss but only a smal percentage.
The rest comes from switching losses when significant amps are switched at voltages. Since switching loss is proportional to switching freqency the faster your CPU the higher the power dissipated.


They have tried to drop the core voltage to reduce this dissipation (and it worked) but at present there is only so low that they can go, and with 3e9 switches per second it doesnt take too many joules in switching loss to dd up FAST!!!

REally they need either SiC processor, but that will just allow the CPU to run hotter = still dissipating power or
diamond processor, switches soo much faster during transition thus reduces stwitchi loss
 
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top