3 ms·
Moore's Law isn't over. You are asking us to believe the recent past means that we will not get more single cpu performance increases over the next 10-20 years.
by melling 11y ago
Moore's Law isn't over. You are asking us to believe the recent past means that we will not get more single cpu performance increases over the next 10-20 years.
And you're saying that a 20%-30% (6x-13x) improvement over then next 10 years won't make a big difference? If Atom was 10x faster, would it be more useable?
As for concurrency not being applicable, that might be true for some tasks but I can easily imagine scenarios where is would help greatly. Multiple buffers and files. Replace and search for example. Indexing, concurrent garbage collectors, ...
So, in your scenario, if cpu's only become 10x faster and we get 60 CPUs, we won't be able to make a better editor?
- coldtea 11y ago>Moore's Law isn't over. The "transistor count" version of Moore's Law has been saved by multiple cores. The speed version hasn't been true for single core performance for years. > You are asking us to believe the recent past means that we will not get more single cpu performance increases over the next 10-20 years.* Moore's law is put in specific time terms -- transistors/power(speed) should double every 18 months. So if this hasn't been the case for the recent 2-3 years, then Moore's Law is already broken -- no belief required. The fact that it's very possible we'll might see another CPU speed breakthrough in, say, 2022 or 2028 or 2050 doesn't change that -- as Moore's Law talks about sustained improvements over 18 month periods. Besides, few in the industry disagree that the Moore Law is pretty much over. That's a great motive behind all the talk of optimizing multicore performance, functional techniques etc. >And you're saying that a 20%-30% (6x-13x) improvement over then next 10 years won't make a big difference? If Atom was 10x faster, would it be more useable? As already said, this 20-30% is "at best" and covers specific benchmarks and uses, not general speed. And nobody said this is sustainable long term either. After a JIT engine reaches a specific point of sophistication, there are few things that one can add to make it even faster.
- falcolas 11y agoCase in Point: Java. Were it possible to make it as fast as C/C++ for the general case, there's been enough money and time thrown at it to get it there. It's fast (frequently fast enough), but it's absolutely plateaued. That Javascript is nearly on par with Java tells me that it's also nearing its own plateau.
- coldtea 11y agoBtw, this just came in: >Next month, the worldwide semiconductor industry will formally acknowledge what has become increasingly obvious to everyone involved: Moore's law, the principle that has powered the information-technology revolution since the 1960s, is nearing its end. http://www.nature.com/news/the-chips-are-down-for-moore-s-law-1.19338?WT.ec_id=NATURE-20160211&spMailingID=50669758&spUserID=NTIwNjgwODkyMTkS1&spJobID=861263427&spReportId=ODYxMjYzNDI3S0 http://www.nature.com/news/the-chips-are-down-for-moore-s-la...