3 ms·
Most power loss in a computer comes from MOSFETs, not resistive loss. Which isn't to say that RTP superconductors wouldn't open up wild new possibilities. ETA:
by scarmig 3y ago
Most power loss in a computer comes from MOSFETs, not resistive loss. Which isn't to say that RTP superconductors wouldn't open up wild new possibilities.
ETA: wrong
- hovering_nox 3y agoMuch of the power consumed, and heat dissipated, by conventional processors comes from moving information between logic elements rather than the actual logic operations. Because superconductors have zero electrical resistance, little energy is required to move bits within the processor. https://en.wikipedia.org/wiki/Superconducting_computing#Fundamental_concepts https://en.wikipedia.org/wiki/Superconducting_computing#Fund...
- amelius 3y agoIncorrect. https://www.ti.com/lit/an/scaa035b/scaa035b.pdf https://www.ti.com/lit/an/scaa035b/scaa035b.pdf This document explains it well. (The resistance of the interconnect is not even mentioned as a significant component of power consumption.)
- XorNot 3y agoYou can synthesize superconducting FETs though: https://link.springer.com/chapter/10.1007/978-94-011-1918-4_8 https://link.springer.com/chapter/10.1007/978-94-011-1918-4_... Obviously this isn't much benefit since the vast majority of applications can't be liquid nitrogen cooled (and computing has followed "what consumers will buy" on progress). That changes substantially if superconductors will keep working when put in my pocket.
- ummonk 3y agoThat statement in Wikipedia is flat out wrong and notably isn't citing any source.
- hatsunearu 3y agoI don't think this is true. Most of the power is dissipated in the metal interconnect.