You must log in or # to comment.
You mean the thing that hasn’t been practically true for two decades?
You don’t say…
It’s not that it hasn’t been practically true so much that it was just hype. Moore wasn’t a scientist, he was a salesman. In practicality, Moore’s law isn’t “we can double the amount of transistors in a chip,” it’s “we can halve the size of transistors.” Which, anyone who’s gone through highschool can tell you, Planck would have something to say about how small you can get. But rephrasing “half the size for the same work” as “double the work for the same size” makes it sound like an infinite growth market, and you know how shareholders love that.
He just observed it, never made some “law” about it. Then it became industry standard, because people want to know what the targets are I guess.
It was never about speed, just constant doubling time for number of widgets per unit of area per unit of cost. That scaling went away a long time ago.
People conveniently forget the ‘per constant unit of cost’ part of the law.
yeah my first thought was we have not made a whole lot of progress in awhile. Until somewhat recently I was using computers that were 10 years old. Its only them dying that got me on a new to me one that is not so old.
Of course it is unsustainable, that’s part of the point. But it shows how progress in computing where a lot of society’s brain has went does continue to impress. Graphics cards have gotten to a point whe D2C is viable and the self reinforcing feedback loop is at play and innovations and ideas how to get more, cheaper compute are tangible.
Yes? So how often can you double CS-3 in the same package? It’s a 5 nm node already, however little that means.
the article is worth a read for the financial aspects of chip fab and mask production and how we will be reaching another plateau in the next decade where more progress is not possible from the economics . we end up requiring some major breakthroughs to push further whereas before we were still squeezing big gains each chip generation in performance, once that ends it changes things . there are a few paths we could take on chips that might workout when all combined but its just as likely we stagnate at that level for decades and software has to stop bloating as well. so the advanced chips are good enough for almost everything practical at this point and there are people making chips at home that are as good as 90s intel chips supposedly , if people use super efficient linux distros its possible to have some sort of neo retro homebrew computer clubs with mesh networking even if the system goes down. , potentially.
