Super flat material could extend life of Moore's Law

[wyattzx]

Quote

Researchers could be fending off the demise of Moore's Law with the help of a new material that allows electrons to move from point A to point B faster. Engineers at the University of Utah discovered a new kind of flat semiconducting material made of tin monoxide that is only one-atom thick, allowing electrical charges to pass through it faster than silicon or other 3D materials.

Source: gizmag, sciencetrib

 

It seems as if every time a major figure in the technology world claims that we're nearing the end of "Moore's Law", something occurs to extend it. Is this a never-ending process for the pursuit of nano technology? Or are we simply stretching the limits, and things like this new material serve a small role in that extension? The article goes on to say:

Quote

Tiwari says the new material fills an important gap in speeding up electronics because, unlike graphene and other near atom-thin materials, it allows both negative electrons and positive charges – or "holes" – to move through it. This has led the team to describe the material as the first stable P-type 2D semiconductor material in existence.

What do you guys think will come of this?

[QueenDemetria]

"Just when you thought you had it all, there was moore." -Moore's other law

[Clanscorpia]

Another "Breakthrough". Wasn't Graphine supposed to be the new 'Big thing'? And I doubt chip makers are going to switch that fast. Those chip fabs cost trillions and they aren't going to upgrade them anytime soon thats for sure.

[flibberdipper]

#Bizwell, since Intel will beat AMD to it lol

 

But I kinda hope this brings more powerful stuff to phones and workstations, for the most part.

 

Hell, low-power stuff for shit like ATMs and checkouts so that we can reduce the amount of power that we waste on those things.

[SurvivorNVL]

Moore.  MOORE!

This is good.

[Dabombinable]

They should make a 40MHz 386 out of it that's compatible with the socket on my 386 mobo (and perhaps a 40MHz 387), then give it to me.

[patrickjp93]

12 hours ago, Clanscorpia said:

Another "Breakthrough". Wasn't Graphine supposed to be the new 'Big thing'? And I doubt chip makers are going to switch that fast. Those chip fabs cost trillions and they aren't going to upgrade them anytime soon thats for sure.

There are many contenders for the next generation of semiconducting materials, but just because it's theoretically possible doesn't mean it's practical financially. You realize all those big etching machines have to be redesigned from the ground up, and the chemical applicators and dissipators have to be as well. Just switching to Silicon-Germanium will be expensive. And getting a pure sheet of graphene right now is next to impossible beyond something the size of a fingernail.

 

tens of billions, not trillions.

[Doobeedoo]

Interesting, good to know that potentially it can extend the current law. Though I'm sure it will pass some time until we see entire new type of processors with new type of materials and all for consumer. But can't wait to see what they'll come up with.

[NumLock21]

So we went from 2d to 3d and back to 2d again.

[Sauron]

The only real thing we need to extend moore's law is give intel some serious competition.

[KemoKa]

4 minutes ago, Sauron said:

The only real thing we need to extend moore's law is give intel some serious competition.

If AMD can take this opportunity, what with Intel making their future processors less powerful, and make chips that are basically the Athlon X2 to their Pentium 4, it'd be awesome and nostalgia would rock the tech world.

[KemoKa]

Also, super-flat material?

I told you that Flat was Justice.

[suicidalfranco]

10 minutes ago, KemoKa said:

Also, super-flat material?

I told you that Flat was Justice.

( ͡° ͜ʖ ͡°)

[captain cactus]

Can we just say bye bye to Moore's law, which is based on a paper released in fucking 1965? How 'bout developing the stuff which is possible at this very time, instead of trying to following something that's over 50 years old at this point and failing at it.

[patrickjp93]

19 minutes ago, Sauron said:

The only real thing we need to extend moore's law is give intel some serious competition.

It really confuses me that you people think Intel lacks competition. There's plenty more money to be made in servers by taking IBM's customers away, and to do that requires being ahead on nodes and delivering world-class tech solutions. Intel competes against IBM, ARM, and Sparc in the server world (and now Huawei is trying to push MIPS_64 as well), it competes against Nvidia in compute accelerators (Knight's Landing vs. Nvidia Teslas, and AMD FirePro as well despite its tiny market presence). It's now also trying to be a leading FPGA producer which requires being ahead on nodes and density, and that is for both enterprise use and to produce the needed logic circuit for every Hybrid Memory Cube module. It's also trying to push innovation in NAND flash and new storage-memory architectures like 3DXPoint. Intel has competition all over the place. Its foundries are pushing as fast as they can. It completely baffles me that you think otherwise.

[Sauron]

Just now, patrickjp93 said:

It really confuses me that you people think Intel lacks competition. There's plenty more money to be made in servers by taking IBM's customers away, and to do that requires being ahead on nodes and delivering world-class tech solutions. Intel competes against IBM, ARM, and Sparc in the server world (and now Huawei is trying to push MIPS_64 as well), it competes against Nvidia in compute accelerators (Knight's Landing vs. Nvidia Teslas, and AMD FirePro as well despite its tiny market presence). It's now also trying to be a leading FPGA producer which requires being ahead on nodes and density, and that is for both enterprise use and to produce the needed logic circuit for every Hybrid Memory Cube module. It's also trying to push innovation in NAND flash and new storage-memory architectures like 3DXPoint. Intel has competition all over the place. Its foundries are pushing as fast as they can. It completely baffles me that you think otherwise.

That's true to an extent, however IBM doesn't really compete directly with intel in the server cpu space. Their business revolves mostly around renting their own megaservers to other businesses. Their pure cpu marketshare is fairly low. ARM and SPARC are in a similar situation, where they mostly sell custom tailored solutions for very specific usage scenarios - not to mention the instruction set differences make them hard to compare. As for competing with nvidia, they are still offering different things that work well in different scenarios - a cpu, even with many cores, is going to have a hard time competing against a gpu in extremely parallelized tasks, whereas a gpu will struggle in tasks where the cpu excels. In a market where people buy hardware on a sufficient scale to have it custom made for them there is room for both, and none can really replace the other.

 

NAND and memory in general are quite separate from cpu architectures and nodes, not all transistors are the same.

 

Although I do suppose it's still in their best interest to stay as far ahead of the curve as possible. I guess the real problem is in the consumer/prosumer cpu market.

[patrickjp93]

24 minutes ago, Sauron said:

That's true to an extent, however IBM doesn't really compete directly with intel in the server cpu space. Their business revolves mostly around renting their own megaservers to other businesses. Their pure cpu marketshare is fairly low. ARM and SPARC are in a similar situation, where they mostly sell custom tailored solutions for very specific usage scenarios - not to mention the instruction set differences make them hard to compare. As for competing with nvidia, they are still offering different things that work well in different scenarios - a cpu, even with many cores, is going to have a hard time competing against a gpu in extremely parallelized tasks, whereas a gpu will struggle in tasks where the cpu excels. In a market where people buy hardware on a sufficient scale to have it custom made for them there is room for both, and none can really replace the other.

 

NAND and memory in general are quite separate from cpu architectures and nodes, not all transistors are the same.

 

Although I do suppose it's still in their best interest to stay as far ahead of the curve as possible. I guess the real problem is in the consumer/prosumer cpu market.

For IBM they have a few main pillars of hardware sales: Wall Street (those financial and market analytics algorithms currently run best on Sparc and Power 8), medical (Watson AI integration), and supercomputers.

 

While true, from a raw compute standpoint Intel still wants to be the king. It needs new nodes for that.

 

in consumerville I'd argue the problem is software, not hardware. We have barely begun the multithreading era even in games. We have the tools to build code that scales to as many cores as we have available, and yet we're still down around 2 threads per game for most, and the usage isn't balanced even for those that go higher or the usage is horribly inefficient.

 

just running a loop like this drives up usage of a core up to 100%.

 

while (true) { //very simple things that take no real horsepower to do; }

[Sauron]

Just now, patrickjp93 said:

in consumerville I'd argue the problem is software, not hardware. We have barely begun the multithreading era even in games. We have the tools to build code that scales to as many cores as we have available, and yet we're still down around 2 threads per game for most, and the usage isn't balanced even for those that go higher or the usage is horribly inefficient.

Yes and no, having 6 or 8 core consumer cpus in the 350$ range would mean hyperthreaded dual cores and no-ht quad cores could come at pentium and i3 prices respectively. Those who don't need more could pay less and those who do need (or simply want) more could get it. If games are the primary use for a computer, wouldn't it be nice to be able to spend 120$ on a cpu and know you won't be seriously bottlenecked in 90% of situations?

[Stefan1024]

Good luck in porducing a 1 atom thinn layer on a 300mm diameter waver.....

But everything to improve silicon is wellcome.

[patrickjp93]

21 minutes ago, Sauron said:

Yes and no, having 6 or 8 core consumer cpus in the 350$ range would mean hyperthreaded dual cores and no-ht quad cores could come at pentium and i3 prices respectively. Those who don't need more could pay less and those who do need (or simply want) more could get it. If games are the primary use for a computer, wouldn't it be nice to be able to spend 120$ on a cpu and know you won't be seriously bottlenecked in 90% of situations?

But the die size and R&D investment, combined with shareholder demands for value, and further combined with the low demand (gaming is a niche use in reality) won't give you the sort of environment needed to get such big chips at cheap prices.

[Sauron]

21 minutes ago, patrickjp93 said:

But the die size and R&D investment, combined with shareholder demands for value, and further combined with the low demand (gaming is a niche use in reality) won't give you the sort of environment needed to get such big chips at cheap prices.

I'm fairly sure production costs go down as the die shrinks - more cpus per wafer means they pay a lot less in resources. R&D would most likely fit in a fraction of what they make at the moment. Of course they act like a business, and they'll try to make as much profit as possible, which is why we need strong competition from amd.

 

I don't think producing a 5960x today is any more expensive than a 2600k was 5 years ago.

[patrickjp93]

Just now, Sauron said:

I'm fairly sure production costs go down as the die shrinks - more cpus per wafer means they pay a lot less in resources. R&D would most likely fit in a fraction of what they make at the moment. Of course they act like a business, and they'll try to make as much profit as possible, which is why we need strong competition from amd.

 

I don't think producing a 5960x today is any more expensive than a 2600k was 5 years ago.

Uh, it's nearly double the die size so... it is more expensive. Mind you the kind of person who needs that horsepower has the money to pay for it.

[Sauron]

Just now, patrickjp93 said:

Uh, it's nearly double the die size so... it is more expensive. Mind you the kind of person who needs that horsepower has the money to pay for it.

you're right, for some reason I thought it was roughly the same size. Still, it could probably be sold at 600 bucks or less at a good profit when you consider a 5820k is made on the same wafers and is sold for far cheaper.

[patrickjp93]

6 minutes ago, Sauron said:

you're right, for some reason I thought it was roughly the same size. Still, it could probably be sold at 600 bucks or less at a good profit when you consider a 5820k is made on the same wafers and is sold for far cheaper.

That's market segmentation for you. And let's not forget AMD launched the 9590 for $1000. Competition does not really guarantee us lower prices.

[Sauron]

45 minutes ago, patrickjp93 said:

That's market segmentation for you. And let's not forget AMD launched the 9590 for $1000. Competition does not really guarantee us lower prices.

not that anyone thought the 9590 was a good offering at launch ^^