/sci/ - Science & Math

Why would that ever happen?

>>9550432
Code monkeys get fired and we go back to custom ASICs to solve tasks.

>>9550435
atoms are only so small.

>>9550450
>atoms are only so small.
Everything is divisible.

>>9550452
Not your intellect, buddy.

>>9550452

We make use of the final frontier
The third dimension

>>9550432
we build better architecture

>>9550716
this right here

>>9550432
they'll get bigger
>>9550459
also this

>>9550716
Well we better start now, any ideas?

parallelism and more cores - GPUs

>>9550432

cpus get bigger

>>9550788

kwandam foam it makes me roam

>>9550432
You make larger chips basically. We will move to anokther medium eventually most likely, lots of research out oon that already.

>>9550432
We find a way to make them smaller even if we have to resort to making processors out of exotic matter. This might occur long, long after CMOS scaling ends. When CMOS scaling ends we can start to consider things like molecular electronics, but that's going to take some time to get working.
>>9550452
Down to the planck length

>>9550432
They become black holes

>>9550905
>planck length
so if electrons arent the charge carriers anymore, what are exactly?

>>9550905
>Down to the planck length
There is no proven physical significance of the Planck length

>>9550432
>What happens when CPU's can't get any smaller?

They stop getting smaller.

We use reversible-comptuting 3D meshes:

Ultimate theoretical models of nanocomputers
http://iopscience.iop.org/article/10.1088/0957-4484/9/3/005/meta

Honestly, why would you ever need to be any smaller than an atom?

We are already trying to make Single-Atom thick passages for electrons to travel down for our smallest currently feasible processors. Last I checked, the problem right now is quantum bullshit making electrons simply pass through the boundaries and into other parts of the chip. We also need more efficient cooling if we want to use CPUs with that kind of gate density.

>>9550432
By smaller you mean the size of the transistor? Then the die will get bigger or multiple dies on one chip like on threadripper

>>9551077
>quantum bullshit
Tunnelling, not bullshit.

>>9550854
cant let the objectively best answer in this thread stay unnoticed

>>9550432
We actually start teaching people the necessity of efficient code and strip away over a century of abstractions in programming

>>9551195
It's a stupid answer. You can't just make CPUs larger and expect them to work faster. A larger die is going to have greater latency due to the distance signals have to travel.

Why do you think processor manufacturers are drifting towards mobile computing? Because they can continue to justify innovation in decreased energy consumption and more compact SoCs that have more 'things' on them.

In the near future we'll see more manufacturers try and develop better all-in-one packages. And in a little while longer we'll have the big breakthrough - 3D-chips that sandwich RAM directly on top of the die. I think the future will bring more specialised CPUs. You already have a special configuration for graphics processing and a general processor but you may have a quantum processor for security applications and more. It will then be a matter of writing programs and system software to make use of the different processors and motherboard manufacturers to make it all efficient.

>>9551251
Cpus could be more than twice their current average size without losing any noticable speed is my understanding. I know im not the best informed but as i understand if a cpu was roughly twice as large as they currently are but the transistors remained the same size, we would get something like 50% additional processing power with no noticable decreases in speed. Maybe im off base a bit and you could lay it out in more detail.

monolithic IC's

>>9550912
You don't need charge carriers to compute. You could potentially compute with colliding neutrons. Of course computing with materials made out of things other than atoms requires exotic matter and is so far off that we should regard such things as science fiction.

Holy shit, people here are fucking retarded if nobody mentioned it yet

P H O T O N I C

I C s

if you're not doing a PhD in Quantum Optics and Optoelectronics right fucking now then you missed the moon mission, buddeh

>>9551204
so will muh python become worthless?

>>9550432
CPU's are 2 dimensional. They are planning on making them a 3-d network to be able to add extra layers. and compact it.

The heat generated would be large, but they will defenitly design coolant systems.

Unless graphene/super conductors are massed produced...


One day anon.... One day...

>>9551616
am I the only one who's impressed by the fact that we can fit billions of transistors onto a tiny silicon chip like that.

>>9551055
Processors get more powerful as their transistors get smaller. The amount we can fit on a processor doubles every tqo years. Eventually we'll reach the limit of how small a transistor can be and not be able make more powerful processors this way.

>>9551463
>Master's includes x-ray scattering optics

Hoo boy it feels good to be winning

>>9551624
This isn't even the end point. I'm sure the future will blow the tits clean off your chest.