[ 3 / biz / cgl / ck / diy / fa / g / ic / jp / lit / sci / tg / vr / vt ] [ index / top / reports / report a bug ] [ 4plebs / archived.moe / rbt ]

Due to resource constraints, /g/ and /tg/ will no longer be archived or available. Other archivers continue to archive these boards.Become a Patron!

/g/ - Technology

View post   

[ Toggle deleted replies ]
File: 2.01 MB, 2592x1936, cutefox.jpg [View same] [iqdb] [saucenao] [google] [report]
43124385 No.43124385 [DELETED]  [Reply] [Original] [archived.moe] [rbt]

What happens to GPU technology when we get down to >1 nm? Can we not get any better?

>> No.43124392

we go to 0.9nm

>> No.43124403


You know what I mean, when we can't get any smaller because we'll be at the atomic level or whatever? Is it even possible to do subatomic stuff?

>> No.43124430

how about just add more cores instead of going lower

that's what nvidia is doing

>> No.43124444

If that every happens (its not) we would measure in pico meters. The smallest possible measure is planks length.

Around 10nm quantum effects like quantum tunneling start become more pronounced because the scale is just that small. Unless there's some sort of breakthrough, chips won't be fabricated below a nano meter anytime in the near future (or possibly your entire lifetime)

>> No.43124449


Does that help performance at all?

>> No.43124616

What a time to be in where we can witness fabricating tech, including CPUs in such small scale. An atom wide within a decade.

Such a shame it'll be with such a shit and old architecture such as x86. We could rebuild a new modern arch from the ground up and make something incredibly good as a new processor for the modern era, like IBM and Fujitsu did.

But nah, that's stupid.

>> No.43125089

But not the power consumption.
You can improve the architecture itself, though.

We will use materials other than silicon, which will decrease power consumption while improving performance (clock speed) even further.

>> No.43125115

Quantum computing

>> No.43125119


It's totally pointless, though.

Modern x86 chips are so fast that it doesn't make jack shit difference for home users and business users are already using different architectures.

Building a new architecture from the ground up is useless when people will just use it to run the same shitty, inefficient software they were running on the old architecture anyway.

>> No.43125139

Assuming it were possible, the smallest silicon transistor would need to be at least 5nm. 1 for each side of the gates, 2 gaps for the electron to travel, and the actual switch in the middle.

It's a lot more complicated than that, but I hope this gives you a basic idea of how a transistor needs to be built.

>> No.43125148

it's impossible even to go to 1 atom thick lines. At least not with sillicon. The material won't behave in a controlled matter anymore. means electric interference, etc.

We are close to hitting that limit. Intel for example already predicted that they won't be able to reduce the size by 2020.

>> No.43125180
File: 190 KB, 800x1000, moarcores.jpg [View same] [iqdb] [saucenao] [google] [report]

Isn't it normally AMD who do that

>> No.43125225

AMD started off with more coars, but it's really starting to catch on with other companies.

>> No.43126830

Instead of making smaller dies and/or adding more cores, they could try to optimize the design to the max...

>> No.43127093

>mfw AMD beats Nvidia with price/performance
>mfw R9 295x2 is cheaper than the titan and is faster

Name (leave empty)
Comment (leave empty)
Password [?]Password used for file deletion.