New Stanford Semiconductor Materials May Take Process Node Manufacturing to 0.6nm

[Rune]

It seems that Stanford may have come across two new materials (hafnium diselenide and zirconium diselenide) that can be used to complement silicon and bypass the 5nm barrier by a considerable amount. It appears that moore's law will continue onward for just a while longer.

 

Quote

As the Stanford researchers started shrinking the diselenides to atomic thinness, they realized that these ultrathin semiconductors share another of silicon’s secret advantages: the energy needed to switch transistors on – a critical step in computing, called the band gap – is in a just-right range. Too low and the circuits leak and become unreliable. Too high and the chip takes too much energy to operate and becomes inefficient. Both materials were in the same optimal range as silicon. All this and the diselenides can also be fashioned into circuits just three atoms thick, or about two-thirds of a nanometer, something silicon cannot do. “Engineers have been unable to make silicon transistors thinner than about five nanometers, before the material properties begin to change in undesirable ways,” Pop said.

Quote

The scientists stress that you'd still need silicon, but the combination of these new materials with silicon could still lead to far more complex processors, much longer battery life, and other advantages that usually come with shrinking transistor sizes. Moore's Law can't last forever (the laws of physics won't allow it), but this could delay the inevitable for many years.

 

Of course, this is going to be many years away if it ever ends up coming to an actual product (and there are other things involved besides just MAKING the physical item). But I like that even though everyone is counting out silicon we are still finding new ways to keep moving forward. At least until we break physics and restart the universe

 

Sources: Engadget ,  Stanford

[RadiatingLight]

How many atoms are in a nanometer?

I was under the impression that it was 2-4.

are you saying we can make transistors that are less than 4 atoms in size?

[mr moose]

1 minute ago, RadiatingLight said:

How many atoms are in a nanometer?

I was under the impression that it was 2-4.

are you saying we can make transistors that are less than 4 atoms in size?

They can switch single atoms on and off in an array, but this is different, and yes they are claiming it "should" be able to go down to 0.5nm I think I read.

[vanished]

Wow, for the longest time I had heard that somewhere around 5 nm would be the limit... that's quite a leap  

 

6 minutes ago, RadiatingLight said:

How many atoms are in a nanometer?

I was under the impression that it was 2-4.

are you saying we can make transistors that are less than 4 atoms in size?

You only need to be concerned once they start claiming we can make them less than 1 atom in size  

[DXMember]

11 minutes ago, RadiatingLight said:

How many atoms are in a nanometer?

I was under the impression that it was 2-4.

are you saying we can make transistors that are less than 4 atoms in size?

14nm process node doesn't mean the transistor is 14nm

the smallest feature is 14nm, the transistor itself can be hundreds of nanometers

and nobody has a standardized way of what they measure, so 14nm GloFo can be bigger or smaller than 14nm Samsung

and when the fabs come out with stuff like Intel's 14nm+ process... what does that even mean?

 

but if I recall correctly someone did estimates that on 5nm process node the transistor gate is supposed to be like 3 silicon atoms wide

[Zodiark1593]

1 minute ago, Ryan_Vickers said:

Wow, for the longest time I had heard that somewhere around 5 nm would be the limit... that's quite a leap  

 

You only need to be concerned once they start claiming we can make them less than 1 atom in size  

Depends how you define the size of the atom. A hydrogen atom is many times smaller than a Uranium atom, soo.

 

[vanished]

Just now, Zodiark1593 said:

Depends how you define the size of the atom. A hydrogen atom is many times smaller than a Uranium atom, soo.

 

Well, atoms of what it's made out of... I would have thought that was fairly obvious  

[DXMember]

2 minutes ago, Ryan_Vickers said:

Wow, for the longest time I had heard that somewhere around 5 nm would be the limit... that's quite a leap  

 

You only need to be concerned once they start claiming we can make them less than 1 atom in size  

we can always use smaller atoms ))) like Carbon

[RadiatingLight]

3 minutes ago, DXMember said:

14nm process node doesn't mean the transistor is 14nm

the smallest feature is 14nm, the transistor itself can be hundreds of nanometers

and nobody has a standardized way of what they measure, so 14nm GloFo can be bigger or smaller than 14nm Samsung

and when the fabs come out with stuff like Intel's 14nm+ process... what does that even mean?

 

but if I recall correctly someone did estimates that at 5nm the transistor gate is supposed to be like 3 silicon atoms wide

Yeah.

the names are really just BS at this point (I think Hynix said that their 14nm process is just a renamed 20nm process with 3D NAND)

however, I'd think that if you're not a big company, like these people at Stanford, then you really need to measure by the actual measurements.

 

seems weird for the gate to be a single atom. actually, seems nearly impossible. but what do I know?

[LienusLateTips]

stop at 0:04

[DXMember]

17 minutes ago, RadiatingLight said:

Yeah.

the names are really just BS at this point (I think Hynix said that their 14nm process is just a renamed 20nm process with 3D NAND)

however, I'd think that if you're not a big company, like these people at Stanford, then you really need to measure by the actual measurements.

 

seems weird for the gate to be a single atom. actually, seems nearly impossible. but what do I know?

mby it's super cooled and running low frequencies so quantum tunneling is not an issue but it's not applicable to mass produced devices either?

[grimreeper132]

20 minutes ago, Ryan_Vickers said:

Wow, for the longest time I had heard that somewhere around 5 nm would be the limit... that's quite a leap  

 

You only need to be concerned once they start claiming we can make them less than 1 atom in size  

I don't know what your talking about, I can't wait until we are using quarks for the material for CPUs, as it will be very strange.

_{(pun intended)}

[RGProductions]

1 minute ago, grimreeper132 said:

I don't know what your talking about, I can't wait until we are using quarks for the material for CPUs, as it will be very strange.

_{(pun intended)}

Nah, we need to first prove the existence of glueballs, then we need to somehow get them to stay, and then we need to make a cpu out of that. That would be interesting.

[vanished]

2 minutes ago, grimreeper132 said:

I don't know what your talking about, I can't wait until we are using quarks for the material for CPUs, as it will be very strange.

_{(pun intended)}

Such a charming product would surely rocket its maker to the top of the market place

[DXMember]

2 minutes ago, Ryan_Vickers said:

Such a charming product would surely rocket its maker to the top of the market place

or get bought out and shut down by some oil company, church or lizard people

[Liltrekkie]

I wonder if this process is vastly different to what current chip makers use, or if it would require a vast retool and investment for not a lot of return. I don't foresee a benefit by going that small in my eyes, is squeezing that 3% more power efficiency worth it? or is there other benefits I'm missing here?

 

Also, not VR ready, will fail. 

[Bananasplit_00]

39 minutes ago, Liltrekkie said:

I wonder if this process is vastly different to what current chip makers use, or if it would require a vast retool and investment for not a lot of return. I don't foresee a benefit by going that small in my eyes, is squeezing that 3% more power efficiency worth it? or is there other benefits I'm missing here?

 

Also, not VR ready, will fail. 

There is probably quite a bit more then 3% to be saved, but even 3% is worth it IMO

[grimreeper132]

6 hours ago, Ryan_Vickers said:

Such a charming product would surely rocket its maker to the top of the market place

you could really say that the marketing for them will be flavoursome

 

_{(yea that one was bad)}

[Juanitology]

seems to me like this is a stopgap measure at best. Graphene transistors are still the future.

[Okjoek]

I was gonna wait until ~late 2018 or 2019 to get a Zen 2 equivalent of the R7 1700 (R7 2700?) and I was basically going to sit on a future-proofed 7nm octa-core CPU for a long time until non-silicon CPUs become a thing which I'm betting will be atleast half a decade. 

 

See any issues with my plans?

[DocSwag]

10 hours ago, RadiatingLight said:

How many atoms are in a nanometer?

I was under the impression that it was 2-4.

are you saying we can make transistors that are less than 4 atoms in size?

You need to properly define what atom. A oganesson atom is gonna be multiple times larger than a hydrogen one. 

[RadiatingLight]

2 hours ago, Okjoek said:

I was gonna wait until ~late 2018 or 2019 to get a Zen 2 equivalent of the R7 1700 (R7 2700?) and I was basically going to sit on a future-proofed 7nm octa-core CPU for a long time until non-silicon CPUs become a thing which I'm betting will be atleast half a decade. 

 

See any issues with my plans?

Yes.

the 2700 won't be an 8-core.

IIRC AMD is increasing CCX size to 6 cores with ZEN+, so it will be a 12 core.

[RadiatingLight]

10 hours ago, Liltrekkie said:

I wonder if this process is vastly different to what current chip makers use, or if it would require a vast retool and investment for not a lot of return. I don't foresee a benefit by going that small in my eyes, is squeezing that 3% more power efficiency worth it? or is there other benefits I'm missing here?

 

Also, not VR ready, will fail. 

 

9 hours ago, Bananasplit_00 said:

There is probably quite a bit more then 3% to be saved, but even 3% is worth it IMO

 

Well intel says that there's 30% less power consumption going from 14 to 10nm, so there are probably hundreds of percents improvement to be had.

[Doobeedoo]

Interesting, can't wait to see it in general use and new process. 

[Okjoek]

4 hours ago, RadiatingLight said:

Yes.

the 2700 won't be an 8-core.

IIRC AMD is increasing CCX size to 6 cores with ZEN+, so it will be a 12 core.

Will it be 300 dollars?