Is there anyway to measure the power consumption of the processor?

[RamPrasadh]

Hi all,



I wanted to measure the power consumed by a processor. Providing a constant supply voltage and a constant frequency for operation,I think that the factor on which the power varies is only the current consumed by it. Suppose If I an running a very complex code, like say image processing stuffs, the overhead of the processor will be maximum. It means , with a constant supply voltage and a constant frequency from the crystal,the only factor which might increase the power consumption is the current flowing to the processor(I am not sure if this is the only one).Is there anyway to measure the current consumed by the processor? If there is one, How do i do that? From where should I probe for the consumed current? The specs have already been given in the data sheets. But how do we measure that manually?There must be a way by which the designers of a processor must have drawn these conclusions on the power ratings of a processor. How do they do it?



Regards

RamPrasadh

 

[T3STY]

For processor you mean a complex CPU like the intel CPUs, or a PIC? Since we're talking about PICs (mainly) in this forum, I'll answer in the case of a PIC (that is still applicable to a CPU, but there might be other parameters to take into account which I could be ignoring).
A simple way to know how much a processor is consuming is to actually measure it at work with a full load (which can even break the component if power consumption is beyond a certain limit).
If that's impossible or you'd like to properly calibrate a power supply before using it, you should consider the case that the oscillator frequency is constant and it's running at the full speed allowed for proper working conditions, the working temperature is constant or it is the maximum temperature that the PIC will be working in, and the output on all pins is constant and at full load (maximum voltage, maximum current allowed). Input pins should not be considered as they are not using current from the PIC power but from an external source, and it does not impact on the power consumption of the PIC itself; any PIC datasheet would usually say the maximum current and voltage (or the power) that a pin can output. If you consider the PIC working in such conditions you'll get a maximum (calculated) amount of power that a PIC could be using. Also, some well made devices can actually allow a bit more power on the output, some other poorly made devices will allow less power on the output; After you've calculated the nominal power consumption you should consider a ±5% to ±10% of error (depending on the accuracy of your calculations).

 

[RamPrasadh]

I will take the case of a PIC itself. If I am not using any peripheral and for instance, if I am performing a Multiply-Accumulate operation , approximately for 10 times, How will I get the consumed current out?
I think I can't extract it from any of the pins.
Is it possible to measure only the power consumption of the core alone? I think the specifications which are given in the data sheets speak only of the power consumption of the core alone on various modes. Correct me if I am wrong.

 

[lucbra]

There must be a way by which the designers of a processor must have drawn these conclusions on the power ratings of a processor. How do they do it?
RamPrasadh

Of course - that's called characterization.
For complex devices (like CPU's, FPGA's, Network processors, ...) it takes months before this process is concluded. And it all starts with simulations and modelling.

And if you want to measure it, build a test PCB, put some testpoints on the PCB so that you can measure the current in all the power supply lines, and you have your numbers.

 

[T3STY]

Is it possible to measure only the power consumption of the core alone?

I guess, there's somehow a method to calculate such power amount too (the best, again, would be measuring it at work). On a PIC16F786 datasheet they say that at 4MHz with 3V power supply the PIC stays under 0.6 mA current consumption (1.8W power). I guess that's the core power consumption only considered at no load.
But why would you need to calculate/measure that? I mean, in such case the PIC would only manipulate some data and output nothing, so it's actually useless because it's not going to control any other device, nor will it tell you any "message" of its elaboration results...

 

[RamPrasadh]

But why would you need to calculate/measure that? I mean, in such case the PIC would only manipulate some data and output nothing, so it's actually useless because it's not going to control any other device, nor will it tell you any "message" of its elaboration results...

I am thinking of taking some work related to this for my masters' degree dissertation. Its like, i want to do something on energy minimization in processors and for that , I am exploring something on the power consumption in processors. It might give some relevant message and the intention is not to make any other device controlled by this. Some suggestions from your side are most welcomed with regard to how can I proceed..

Of course - that's called characterization.

I will look on to that and revert back.