Quantum limit for micro-architecture?

[KezzaMcFezza]

Hey guys, so i heard somewhere last year that CPU's and GPU's and even SSD's aren't going to shrink much longer. If i remember rightly he said that moores law will l stop at 7nm. Is this true? If so is it possible to break physics if we got smaller?

 

Thanks in advance

[Lurick]

With current silicon based technology I think the limit is 7nm maybe 5 but we can easily go smaller with things such as nanotubes and non-silicon based processors.

[Moonzy]

yeap, quantum physics is a bish

 

but im sure the geniuses will find a way around it

[KezzaMcFezza]

Just now, Lurick said:

With current silicon based technology I think the limit is 7nm maybe 5 but we can easily go smaller with things such as nanotubes and non-silicon based processors.

Ahh, so maybe we can have carbon nano-tube CPU's soon?

[mikat]

Just now, ProKeero said:

Ahh, so maybe we can have carbon nano-tube CPU's soon?

not soon, but in 10 years we will know more

quote me in 10 years when those things are around (maybe we have galium arsenide transistors in 10 years idk)

[KezzaMcFezza]

1 minute ago, Moonzy said:

yeap, quantum physics is a bish

 

but im sure the geniuses will find a way around it

Maybe we could have massive CPU's? like ones as big as a GPU?

[Moonzy]

Just now, ProKeero said:

Maybe we could have massive CPU's? like ones as big as a GPU?

we already do

[KezzaMcFezza]

Just now, mikat said:

not soon, but in 10 years we will know more

quote me in 10 years when those things are around (maybe we have galium arsenide transistors in 10 years idk)

Lol in 10 years my i5 4690k will be like the Core 2 Duo of today

[KezzaMcFezza]

1 minute ago, Moonzy said:

we already do

What the actual fuck..  why hasn't linus done a video with 2 of these yet

[mikat]

Just now, ninninon said:

I can't even contemplate how small 45nm is... Electronics confuse me enough as it is, but the fact that they're fitting all these magic things on sand amazes me....

about 450 hydrogen atoms thick lol, that's really tiny

[TheMcSame]

-EDIT-

NVM just me derping.

[mikat]

4 minutes ago, ProKeero said:

Maybe we could have massive CPU's? like ones as big as a GPU?

the gpu on a graphics card isn't much bigger or bigger at all compared to a cpu, a graphics card is: memory, board and gpu.  (that's why graphics cards are so large compared to a cpu)

[KezzaMcFezza]

2 minutes ago, ninninon said:

Theses must be some seriously small magic fairies they've got. Maybe they're hiding amputee fairies inside them. 

#QuadraplegicFairiesLivesMatter

I don't really get how they can manufacture something as small as 14nm  or even 88nm for all that matter. What tools would you use? lol

[agent_x007]

7nm is considered a limit for Silicone (Si), because it's atoms are 0,21nm in diameter (Van der Waal).
To make it simpler to understand :
Have U tried to build a waterproof wall (gate), with a few dozens of ping-pong balls (round shape objects) ?
That's more-less how hard it is, to make 7nm transistors.

[Coaxialgamer]

Yeah , silicon-based substrates will top out at around 7nm ( maybe up to 5nm ibut that would be pushing it ) . It seems we will need to shft technologies to go forward .

Here are the main possibilities we have :

-quantum computing ( the elephant in the room) : the workings of quantum processors are fairly complicated , and i won't explain the here , but i don't think quantum computing will become mainstream because of it's very nature . Apart from very specific tasks , it shouldn't  be faster than traditionnal computing . The main problem with this is the extreme temperatures required to run .

 

-graphene transistors ( commonly know as CARBON NANOTUBES TRANSISTORS ): graphene requires much less energy than silicon thanks to greater electron mobility , and allows frequencies much higher than traditionnal cpus ( graphene transistors can go up to 1000ghz ): the main problem with this is that we can't manufacture it on a large scale yet.

 

-ternary computing ( my favourite ) : in theory, it's possible . We would use ternary logic instead of binary ( in fact , ternary computers have been made ). the main advantage here is that we need less transistors to express the same value ( 64 bits represents the same value as ~40 "ternary bits") . This means we need less transistors to do the same work , saving energy and cost ( allowing to pack even more power into chips ) . But as with binary logic , we would reach quantum limits again eventually . the main drawback here is obvious : this makes circuits MUCH more complicated , and breaks compatibility with almost alll code ever written 

 

-biological computing : using proteins, chemical signals etc , we represent data .  This is a very viable alternative to quantum computing while being cheaper . As this is very new , power and speed potentials are unknow , but scientist think it's a good alternative to classical computing .Main drawback : VERY new.

 

-optical computing : using photons to represent data .This would increase computing power exponentially , while significantly reducing power and heat. Problem is , it's very new , and there are problems with data leaking out of the processor ( it's light , so duh ). Apart from there being no fully functionning prototype out yet , we don't really know how to manufacture it .

 

- use different substrates materials like GaN ( galluim nitirde ) and gallium arsenide . They both have their benefits ( lower power , faster frequencies ), but GaN has a signifcant advantage : it can be manufactured using existing fabs . This is a huge plus , and companies are more likely to adopt it . The main problem : only one company is currently really putting work into it , and there is the problem of existing silicon manufacturers not being very happy about it.

 

Edit : i forgot stacking , but that relies on existing technolgies ( think of it as 3D nand for processors ) . The main problem here is heat dissipation

[Oberon.Smite]

27 minutes ago, Moonzy said:

we already do

This is more of a secondary/coprocessor rather than an actual CPU. Mostly used for crunching numbers. It technically counts, but I don't think we have any CPUs that have an actual die the size of a GPU.

[Coaxialgamer]

1 minute ago, AlexTheRose said:

Not true. The most I’d gander it breaking compatibility with is processor microcode, because the higher you go the more abstractions you encounter with “code.” CPU instructions themselves could be reimplemented with no compatibility break whatsoever, so there’s no real threat of broken compat outside of the research lab for this technology.

 

I’ve worked with ARM assembly in the past and while I don’t know volumes about ternary computing I’d say the functionality of machine code is transferrable without issue. And even if it’s not, we can always jump the abstraction chain and go up a link.

but circuitry would be much more complex though , because of 2 bits = 4 possibilities , it would be 2 "trits" = 9 possibilities , and it gets exponentially worse . think of the truth tables ! What a nightmare for circuit designers !

[Moonzy]

25 minutes ago, ProKeero said:

I don't really get how they can manufacture something as small as 14nm  or even 88nm for all that matter. What tools would you use? lol

 

basically, a kid burning an ant hill with a magnifying glass

[juretrn]

37 minutes ago, ProKeero said:

Ahh, so maybe we can have carbon nano-tube CPU's soon?

CNTFET transistors have been made in the lab already...

but not anywhere close to mass production needed to make chips.

For now, graphene is probably the best bet AFAIK.

[Coaxialgamer]

29 minutes ago, mikat said:

the gpu on a graphics card isn't much bigger or bigger at all compared to a cpu, a graphics card is: memory, board and gpu.  (that's why graphics cards are so large compared to a cpu)

 

31 minutes ago, ProKeero said:

What the actual fuck..  why hasn't linus done a video with 2 of these yet

the main issue here is yields : the larger the chips , the less you get per wafer . while it is true that the largest cpu's AND gpus are around 500-600mm² , cpus tend to sell with much higher margins. , and therefor , tend to be smaller on the consumer side. That and since the cpu market is much bigger , bigger volumes of them are needed.

[laminutederire]

Carbon based processor will probably be the best option to get really powerful stuffs very fast. Why? Because they figured out the electronic properties of few atoms thick are basically amazing. This would allow high number of transistors, and huge clock frequencies thanks to the fact it won't heat that much and that circuits will handle it.

Eventually we'll have to have fun with quantum electrodynamics to push circuits to amazing frequencies more than trying to push out the number of these.

[xentropa]

Spin based computing might be the future.

 

https://en.wikipedia.org/wiki/Spintronics