1nm GPU chip?

[Simon771]

So if I understand right, the lower the number of "nm", the lower temps we get and that can also mean higher clocks?

 

If it keeps going like this, we can expect to see 1nm GPU card in like 10 years? It will be able to have 8192 cuda cores at 4500MHz with passive cooling

 

Well one can hope so right?

 

 

I would also like to ask what does that "nm" even means, and if you guys think that 1nm cards could realy be a thing.

[WereCat]

Uhh..

nm - nano meter (size of manufacturing process)

 

And no ... I am sure that we will not see any 1nm chips within 10 years. If that is even possible to manufacture (I know it is not possible to manufacture that on silicon).

 

And shrinking down manufacturing process wil only get you so far. More important are newer technologies like graphene for example.

[Chewique]

A nanometer is a unit of spatial measurement that is 10-9 meter, or one billionth of ameter. It is commonly used in nanotechnology, the building of extremely small machines

[Enderman]

no

silicon atoms are too big

 

nm is a measurement of distance and how small something is

its called nanometers

and no, smaller =/= lower temps + higher overclocks

it depends on a lot more than just nm

[Simon771]

1 minute ago, WereCat said:

Uhh..

nm - nano meter (size of manufacturing process)

 

And no ... I am sure that we will not see any 1nm chips within 10 years. If that is even possible to manufacture (I know it is not possible to manufacture that on silicon).

What's the lowest they can go on silicon?

[ZetZet]

1nm not doable on silicon. 5nm is thought to be the absolute minimum. But there are 3nm prototypes out of couple of transistors.

 

 

Silicon atom is 0.111nm so as you can imagine, going down to just a couple of atoms of width isn't very easy.

[WereCat]

Just now, Simon771 said:

What's the lowest they can go on silicon?

It was supposed to end at 10nm ... looks like they were wrong and 7nm is the lowest we will see. That said there are different physical laws that apply at these sizes and they have to deal with quantum tuneling... you can google these thing easily.

[ThinkWithPortals]

1 minute ago, Simon771 said:

What's the lowest they can go on silicon?

About 5nm. Lower than that and electrons can skip, making the chip useless.

 

A new material is being looked into that could go smaller than 5, however: graphene.

[ZetZet]

2 minutes ago, ThinkWithPortals said:

About 5nm. Lower than that and electrons can skip, making the chip useless.

Not only that but they could physically rip the paths apart because of how "big" they would be.

[Arc_Jester]

That is not possible with silicon, in order to possibly get that small carbon nanotubes or something similar will have to be used in place of silicon.

[ThinkWithPortals]

Just now, ZetZet said:

Not only that but they could physically rip the paths apart because of how "big" they would be.

True also.

[laminutederire]

3 minutes ago, Simon771 said:

What's the lowest they can go on silicon?

Something between 5 and 7 nm. You have to understand that 0.1-1nm is the order of magnitude of an atom's size. That being said, they have still a tough way to achieve the limit, and there will be even more powerful things in the future which won't heat that much thanks to better electrical conductivity, and therefore less heat dissipation. If you want yo understand those things, I'd be happy to give you some references to learn about it.

[Atmos]

1 minute ago, WereCat said:

Uhh..

nm - nano meter (size of manufacturing process)

 

And no ... I am sure that we will not see any 1nm chips within 10 years. If that is even possible to manufacture (I know it is not possible to manufacture that on silicon).

100% not possible on silicon.

also 100% not possible in 10 years so you were bang on with those points.

 

 

As for the op, it becomes much more difficult to develop designs and manufacturing techniques the smaller the die size becomes. For 1nm to become a reality we would first need a new material to etch it onto, and a system of etching far.................. FAR more advanced than what we have. 1nm isn't impossible, but with our current rate of development it will take us approximately 15-20 years to see any sort of viability. 

Unless of course some alien spacecraft lands in the middle of L.A. and they go all "join the galactic federation, and shit" on us. Or if we develop a self-learning. self-upgrading AI for the sake of technological design. Either of the two will do, but I'll be honest, option 1 is far better imo.

[Simon771]

well that was a lot of replies in such short time

 

Googled about graphene and that seems like cool future technology.

 

So I guess we will be able to see 12nm, 10nm and 7nm GPUs with current technology in years that are yet to come.

 

Also someone said that lower nm doesn't mean less heat. Well I don't know what magic they did at Nvidia, but 2100MHz on 2560 cores with only 1 little fan and temps won't go above 67°C ? Well that's just awesome. Guess it have to do something with power effiecenity so if it uses less power, it produces less heat. But still ... that's impressive.

 

[ivan134]

3 minutes ago, Simon771 said:

well that was a lot of replies in such short time

 

Googled about graphene and that seems like cool future technology.

 

So I guess we will be able to see 12nm, 10nm and 7nm GPUs with current technology in years that are yet to come.

 

Also someone said that lower nm doesn't mean less heat. Well I don't know what magic they did at Nvidia, but 2100MHz on 2560 cores with only 1 little fan and temps won't go above 67°C ? Well that's just awesome. Guess it have to do something with power effiecenity so if it uses less power, it produces less heat. But still ... that's impressive.

 

Well yes and no. It means less heat at the same clock speeds, but then we just bump up clock speeds to make things faster, so in the end we end up with the same temps.

 

Smaller node = less distance electrons have to travel = less energy spent = less heat generated

[Simon771]

4 minutes ago, ivan134 said:

Well yes and no. It means less heat at the same clock speeds, but then we just bump up clock speeds to make things faster, so in the end we end up with the same temps.

 

Smaller node = less distance electrons have to travel = less energy spent = less heat generated

Exactly what I was thinking before.

Imagine that 1080 running at 1200MHz or 1100MHz like titan x or something. Temps would be under 50°C probbably  

[Godbert Mason]

On 5/11/2016 at 5:50 PM, Simon771 said:

So if I understand right, the lower the number of "nm", the lower temps we get and that can also mean higher clocks?

 

If it keeps going like this, we can expect to see 1nm GPU card in like 10 years? It will be able to have 8192 cuda cores at 4500MHz with passive cooling

 

Well one can hope so right?

 

 

I would also like to ask what does that "nm" even means, and if you guys think that 1nm cards could realy be a thing.

A 1nm gpu??..well that would have been great...but the laws of physics wouldn't allow this...we wouldn't want electrons jumping from one transistor to the other...thus causing wrong data "compution"...PS:yh I know...it's not English..