No. If you’re thinking Moore’s Law, all that says is that semiconductors would double their number of transistors every 2 years (or so) for a given price point. This is basically making a really big package while keeping the price the same.
BTW, that price limit is one that isn’t really talked about, but it’s in Moore’s original paper (unlike things people usually bring up, like clock rates or single threaded speed), and it skewers Moore’s Law dead. If you take the price of the old 8086, adjust for inflation, and double its transistors every 2 years since, there’s nothing that comes close to the numbers you get. IIRC, it’s about an order of magnitude too few transistors for CPUs on offer at that price.
I think the point they were wondering was that a larger computer chip doesn’t seem like progress. The overall density of transistors is the same, so how exactly does scaling that up do anything? Or why does using glass make it better?
Granted, reading the article answers exactly that (though I’ll admit, I don’t entirely understand it). The current material limits how much of the computer chip somehow, this new material allows for more… something.
The article is talking about how they make the boards that cpu chiplets are placed on. The chiplets themselves have technology that is getting smaller. The point of an upgrade in the board is to fit more of these chiplets in a device, so you can process more information in parallel. That’s why they’re focusing on AI computing which relies heavily on parallel processing.
The issue is that chip sizes are limited (among other things, like the speed of light) by the ability to take away the heat they produce. For example the i9 processor itself is not that big, but needs a massive cooler.
Bigger? Aren’t they supposed to get smaller?
No. If you’re thinking Moore’s Law, all that says is that semiconductors would double their number of transistors every 2 years (or so) for a given price point. This is basically making a really big package while keeping the price the same.
BTW, that price limit is one that isn’t really talked about, but it’s in Moore’s original paper (unlike things people usually bring up, like clock rates or single threaded speed), and it skewers Moore’s Law dead. If you take the price of the old 8086, adjust for inflation, and double its transistors every 2 years since, there’s nothing that comes close to the numbers you get. IIRC, it’s about an order of magnitude too few transistors for CPUs on offer at that price.
I think the point they were wondering was that a larger computer chip doesn’t seem like progress. The overall density of transistors is the same, so how exactly does scaling that up do anything? Or why does using glass make it better?
Granted, reading the article answers exactly that (though I’ll admit, I don’t entirely understand it). The current material limits how much of the computer chip somehow, this new material allows for more… something.
I wont rest until my CPU is delivered to my house on a pallet, just as god intended.
Longing for the day when I’ll finally get to ask my friend if I can borrow his 18-wheeler so I can haul home the CPU I bought on craigslist.
The article is talking about how they make the boards that cpu chiplets are placed on. The chiplets themselves have technology that is getting smaller. The point of an upgrade in the board is to fit more of these chiplets in a device, so you can process more information in parallel. That’s why they’re focusing on AI computing which relies heavily on parallel processing.
The issue is that chip sizes are limited (among other things, like the speed of light) by the ability to take away the heat they produce. For example the i9 processor itself is not that big, but needs a massive cooler.