Energy Per CPU cycle

1) measure current and voltage 2) calculate consumed energy (per second) 3) divide by CPU frequency (clocks per second).

Ilmars said:

1) measure current and voltage 2) calculate consumed energy (per second) 3) divide by CPU frequency (clocks per second).

The CPU is not the only part that requires power. There are several things that require a static power supply and are not scaped with PCU frequency.

Also, of course the result will only be an aerage. An isntruction that takes one CPU cycle but only works on internal registers, is different from an instruction that has to read offset values and parameters from ram or flash.

Also, if the code is running from ram, this is mroe efficient than when running from flash.

On the bottom line, there is no single specific 'power per CPU cycle' value.

Jens-Michael Gross said:

On the bottom line, there is no single specific 'power per CPU cycle' value.

Because execution of each specific cpu instruction can take specific amount of power. Main question here is - what's the point to know it? Just to keep us busy? :D

I'm calculating (approximate) CPU consumption as follows:

• set SMCLK to DCOCLK/8
• connect TACLK to SMCLK
• TB is counting time (with low frequency)
• on each overflow of TA one knows that there were 8*65536 CPU cycles (sleep times (i.e. LPM3) are not counted)
• TA is running in continous mode
• from the data sheet one knows that the CPU needs x[mA/MHz]

With the information of "TA overflows" (=AM time), "TB clock" (=total time) and the datasheet it is relatively easy to calculate actual consumption.

Hardy

Ilmars said:

what's the point to know it? Just to keep us busy?

But as we can see every day in this forum (and also in the Windows world), the most efficient processor is worthless when it is tortured by an unskilled coder.