CPU power consumed relates to information processed?

[Delta2]

As far as i know CPUs consume power because of the ohmic resistance of the transistors that they are built of(i hope i am not wrong on this).

But if we were able to build some sort of superconducting CPU with almost zero ohmic resistances, would it be able to operate and perform computations with almost zero power, or (and here comes the interesting part of my question) because there are computations and hence information transformed, the entropy of the CPU is changing hence there has to be some non neglible power consumed by the CPU(even if it was superconducting)?

 

[rumborak]

https://en.wikipedia.org/wiki/Landauer's_principle

 

[Delta2]

So, if the CPU has zero ohmic resistances the CPU can operate fully with really very small power consumed as dictated by the Landauer's Limit?

Why i do feel something is not right, i feel like we are getting something (information processed by the CPU) while offering nothing or almost nothing as Landauer's energy is really small even if information is allowed to be erased?

 

[CWatters]

CMOS devices have capacitance (both the traces on the chip and the gate capacitance of FETs). This has to be charged and discharged when the logic level/voltage changes.

 

[sophiecentaur]

CMOS devices have capacitance (both the traces on the chip and the gate capacitance of FETs). This has to be charged and discharged when the logic level/voltage changes.

In an attempt to increase speed without increasing the Energy dissipated per cycle, chip makers have been using lower and lower supply voltages (E = CV²/2). This, of course, brings in other problems due to the low signal levels.

 

[mheslep]

As far as i know CPUs consume power because of the ohmic resistance of the transistors that they are built of(i hope i am not wrong on this).

But if we were able to build some sort of superconducting CPU with almost zero ohmic resistances, would it be able to operate and perform computations with almost zero power, or (and here comes the interesting part of my question) because there are computations and hence information transformed, the entropy of the CPU is changing hence there has to be some non neglible power consumed by the CPU(even if it was superconducting)?

I think Landauer's given above is the applicable principal of information processing, but here are the particulars of why information as manipulated by transistors must draw non-zero power.

Electronic computers represent binary information by means of transistors in an "on" or "off" state, i.e. a switch. The on state is characterized by relatively low resistance, and indeed transistors are constantly improved by further lowering the "on" resistance. But the "off" state is defined by relatively high resistance, the as switches are, and the opposite of the point of superconductors. To prevent loss in the off state the resistance should be infinite, but it never quite is and so when "off" transistors "leak". They leak less and less apparently with each new development but it never completely goes away. Furthermore, even though the switching time between on and off has decreased by many, many orders of magnitude with ever-new semiconductor developments, that time remains, and will remain non-zero. It is during this time that some current flows through a middling resistance and creates ohmic losses. There are still more imperfections in the model of transistor as a perfect all-on/all-off switch.