WyomingKnott
Legenda in Aeternum
ptmmac :
Yawn, the death of Denson scaling has made each new CPU introduction so boring. Yes I would love to have better battery life, but when it takes decades to get significant results it is pretty hard to get very excited about things. During the late 90's we saw 1000 fold improvement in processing speed for single threads. We need a new process based on light that is not so close to melting down from power loss.
I think that the focus has shifted from massive single-thread power to efficiency per watt of power; that keeps increasing. For applications which will take decades to get an answer in a single-threaded process, the best solution (if it can be applied to the current algorithm or if an appropriate new algorithm can be made) is to massively parallelize the process. At which point getting more gigaflops for your watt of power becomes a blessing.
Out of curiosity, not as a personal challenge to your post, can you name major problems which still have to remain single-threaded to run on computers?
I do agree with you on the point that optronics, or whatever it will be called, will probably completely shatter the current instructions-per-second-per-watt barrier.