Discussion
Moore's Law is dead.
Intel introduced the 64 bit "Core 2" processor in 2005 so laptops could access more than 4 gigabytes of RAM.
Moore's Law's 18 month doubling time says the following 21 years should have 14 doublings to 64 terabytes of RAM in each $2000 laptop.
That's why silicon valley execs went crazy: they no longer have free growth. Their business model doesn't WORK. They've done regulatory capture, brute force spending inflating a snake oil bubble, and nothing's working.
@landley
Moore's law is indeed largely gone. But it was about transistor density, not about the capacity or price (Moore never claimed things would get cheaper!) of the resulting systems, and it never really applied to ram.
@jannem The original 1965 Moore's Law paper was about RAM. Intel was founded to produce DRAM.
https://www.landley.net/history/mirror/interviews/Moore.html
RAM doubled like clockwork (largely due to industry collusion in production schedules) for decades, I wrote about it in 2003:
http://www.catb.org/~esr/halloween/halloween9.html#id2867629
And again in 2006, that time with example products/ranges sourced from an extensive collection of old computer magazines:
http://catb.org/~esr/writings/world-domination/world-domination-201.html#id248066
@landley
I stand corrected - he was talking about ic components in heneral! And I agree Moore's law no longer fully holds (and will eventually slow to a crawl).
But my greater point is still valid: it talks about manufacturing density. Not about the capacity, speed, or price of the resulting product. The end of Dennard scaling already decoupled that. Smaller components aren't able to run faster, or made cheaper.
@jannem Chip designs hit a clock speed wall circa y2k, then went wide (pentium M was basically just a pentium III with the racy logic cleaned up and megabytes of L2 cache on die), then they went SMP, then the put frame buffers and eventually GPUs on die to soak up the transistor budget (AMD bought ATI in 2006).
The mainframe->minicomputer->microcomputer->smartphone migrations muddied the waters a lot (I've given talks about those), but when I say moore's law is dead past tense, I mean it.
@landley
Yep, that was the end of Dennard scaling. I think people underestimate the impact of that.
Your RAM observation got me thinking:
It's easy to forget that Moore's law was an economic observation. It's says it's worth keeping that development pace, because the revenues will make it worth the cost.
Both sides of that equation have to hold. Each new process node costs more. So the benefits side has to increase in proportion. And that's stopped happening for two reasons:
🧵
I'm typing this on a laptop that came out in 2012. It was high-end then and I've upgraded it as far as it can go, but no piece of hardware that's NINE DOUBLINGS behind should still be usable, let alone slightly more powerful than a modern chromebook.
That's why silicon valley's gone septic. Everything they did was pantomime while photolithigrapic die size shrinks did all the heavy lifting. Now the S-curve's bent back down again there's no more growth to be had, and they're in SCREAMING DENIAL.
Install bonfire.cafe
Get the full app experience
