DARPA, PHYS.ORG, DARPA circuit achieves speeds of 1 trillion cycles per second, earns Guinness world record, here.
Officials from Guinness World Records today recognized DARPA’s Terahertz Electronics program for creating the fastest solid-state amplifier integrated circuit ever measured. The ten-stage common-source amplifier operates at a speed of one terahertz (1012 GHz), or one trillion cycles per second—150 billion cycles faster than the existing world record of 850 gigahertz set in 2012.
Robert W. Lucky, IEEE Spectrum, Has Moore’s Law Become Moot for PC? here. He sounds like a High Frequency Trader: “my gcc code runs about the same speed as on Westmere, can’t Moore’s Law make the context switching really fast?” I think Pope Frankie answered this when he noted that God is not a magician with a wand. I think what the Pope is trying to say here is, even with ever exponentially improving Moore’s Law, clueless code will continue to run like, wait for it, … clueless code. There was a time when dopey code got better performance with new silicon. That ended like 10 years ago.
Depending on which benchmark programs I run, the actual observed factor of improvement in performance between the old and new systems varies from about two to six. However, this brings me to my second thought: Unless I’m running benchmark programs, I don’t notice any difference in performance. Most everything I do is either so fast anyway that I don’t notice a difference or is constrained by some other bottleneck, like Internet connectivity. Of course, if I were a heavy user of games, the difference might be evident, but it would depend mostly on the graphics card used.