Moore's Law no longer happening

[jbMacAZ]

Moore's Law seems to hold for the systems I can afford. I can roughly double performance for about the same cost every two years. I can't afford the bleeding edge machines where Moore's Law no longer holds.

[entropyfoe]

Single thread speeds are topping out. My signature machine is 3.6 GHz 12 threads, and the super computer I replaced with it was also at 3.6 GHz but 8 threads. My wife's Wind 7 machine was 4 cores/4 threads at 3.7 GHz , so in the past 5 years or so, not much progress on clock speeds, (mostly due to heat dissipation) only adding cores, and a bit of instruction per clock improvements. Sure you can turbo up your CPU or over clock to 4 Ghz and beyond, but heat is the usual limitation.

Os unless you and your application can use more cores, more multi-tasking is all you get. (like the massive AMD threadripper now available 32 cores 64 threads) And the fabrication technology stagnating, waiting for the next big thing (graphene or what ever it will be) with my Ryzen last year coming in a 14nm process, same as the latest Intel. This year AMD did a "shrink", same design going from 14 -> 12nm and Intel is stuck with problems bringing their 10nm into production. Most companies cannot justify a 250 million dollar mask set and design, only a few high volume, high value designs from the Intel/TSMC/Samsung club will viable.

So progress on the hardware side will slow down, and I hope Linus is right, software will have to pick up the torch to advance computation at the steady pace we have been spoiled by. We are in a golden age, enjoy your computers, where a 5 year old machine is totally fine for most users, creating a surplus of machines for linux to be installed on.

For me the biggest boost in the past few years has been storage, with the rise of SATA SSDs, the blindingly fast M2 PCIE 3.0 x 4 drives, and massive 4-10TB rotating drives.

[KBD]

entropyfoe, I believe you are correct that this is the golden age of Linux. So much hardware works out of the box with Linux and so well with Linux now.

[Pierre]

there was a massive increase in Internet Speeds, when dial-up was finally replaced,
but now, there is a stabilizing in those Broadband speeds.
- further more, like in this country, the Broadband is being stuffed up by Politicians Et Al.

but, it's still nice to be able to resurrect an Old Vista Based Machine & chuck a copy of MX onto it,
& watch it start to Fly Again, since the difference between those machines, and the current crop of machines,
is not as big a gap as there once was.

things like a New SSD will make more of an difference, than any gains in CPU / GPU or even in memory increases.

[KBD]

I don't think we are anywhere near the end of increases in broadband here in the U.S. It might top out at Gigabit speed in large cities, but too many in rural areas will be waiting years, if then, to hit those highs. It generally doesn't cost much more for most ISP's to increase speeds, and it gives them an excuse to increase prices with "up to" speed promises. I live in a modest sized city with 2 ISP choices--which makes me very lucky--I'm not limited to one expensive crappy option. A local ISP ran fiber optic lines in my part of the city and we get Very Reliable 25 down and 5 up speeds. Spectrum is our other choice and the only way I would have Spectrum would be if they were my only choice.
I expect our speeds to increase in the next few years. Right now 25/5 is good, especially as I actually get the promised speeds and outages can be counted on one hand over the past 3 years.

[Gordon Cooper]

An article in BBC Business News : https://www.bbc.com/news/business-45273584
discussing Quantum Computers contains this comment:

" Moore's Law - that every two years, microprocessors will get twice as fast, use half as much energy, and take up half as much space - is finally breaking down."

[BillZM]

To echo above comments, the "need for speed" seems to still drive progress, but the processing bottleneck on which primary efforts/progress are focused has shifted, I think, from PC hardware to broadband to phone/tablet to cellular networks. While these latter components catch up, PC hardware capability has been more than sufficient for everyday tasks and thus trending more toward lower cost than greater capability. As mentioned, lighter weight Linux distros and SSDs enable even greater useful life from our aging PCs. That's not to say there aren't more and more capable microprocessors and PCs being introduced. Maybe we should look at consistent price points, say $1500 desktops or laptops, or $500 microprocessors, to gauge ongoing Moore's Law progress.

[Pierre]

that $1500 price mark is what I'd paid for my first ever machine - - Z80 based ..
and the price that I'd paid for my second PC based machine,
so that is a good benchmark - - sure, you can go even lower & much higher, as well,
but that is a good benchmark for today's machines.

[KBD]

It is interesting to me that they are actually making faster chips, even faster mobile chips, but the infatuation with 'thinner and lighter' is killing performance. I would actually prefer a heavier device that had more power than a feather-lite device that is encumbered by heat and a crippled cpu due to heat.

[clicktician]

It is interesting to me that they are actually making faster chips, even faster mobile chips, but the infatuation with 'thinner and lighter' is killing performance.

Gordon Moore's law is only about miniaturization. He predicted in 1965 that the number of transistors on a silicon chip would double every year (and later lengthened that to 18 months). He didn't make any predictions about speed or power. Only density. Then we invented the whole performance-doubling thing and attributed it to him.