What is after 1nm?

[Lord Szechenyi]

4 hours ago, baxuz said:

what comes after 1nm

your mom

 

i'm sorry i had to

[whm1974]

12 minutes ago, Drama Lama said:

same was said about normal computers

 

We needed the Microprocessor to be invented first. Althrough a Microcomputer could have developed using TTL Dies or Bitslices.

https://en.wikipedia.org/wiki/Transistor–transistor_logic

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

 

The Datapoint 2200 was suppose to use the i8008 Microprocessor, but the Company that hired Intel to developed it, instead used TTL and Bit Slicing dies and allowed Intel it use then paying the Fee.

https://en.wikipedia.org/wiki/Datapoint_2200

[Mark Kaine]

6 hours ago, Drama Lama said:

same was said about normal computers

 

yeah but only because they were several hundred square meters, probably spanning several buildings!

[Spindel]

13 hours ago, Drama Lama said:

x nm doesn't mean anything it's not that a 7nm chip is exactly 7nm big there are a lot of factors other than transistor size like density and the transistors themselves but yes at some point you can't make a transistor smaller but still increase transistor density ( " having less room between transistors " ). and existing nodes can be made better like 14nm +++ and 5nm+

it was once said " well you can't go past 100nm and now we're discussing about 5, 3 and 2nm process nodes.

Thats why I said the switch will be to silicon carbide. 

Higher density (maybe with 3d layout of transistors) will cause thermal issues with current materials. A material like silicon carbide (SC) can operate at much higher temperatures, which will mitigate these problems somewhat since the problem will be to get heat out of the middle of a layer design. Also SC can allow for higher frequencies because of this. 

 

SC is currently used in high power transformers, the hope is it will make it to computer semi conductors 

[Drama Lama]

Just now, Spindel said:

can operate at much higher temperatures,

Yes but people don’t want phones to be hot

[Letgomyleghoe]

16 hours ago, whm1974 said:

If I'm not mistaken Silicon Semiconductors are limited to 5nm size for the Transistors. That wouldn't stop continuing improvements 5nm Nodes once we reach there. After all just look at what Intel manages to do with 14nm.

 

 

less than that theoretically they could get down to 1 atom (I don't remember how big they are but I think 14nm is about 7 atoms, so that would be like 0.5nm max on silicon, but then we could always move to graphene to so who knows

 

27 minutes ago, Spindel said:

Thats why I said the switch will be to silicon carbide. 

Higher density (maybe with 3d layout of transistors) will cause thermal issues with current materials. A material like silicon carbide (SC) can operate at much higher temperatures, which will mitigate these problems somewhat since the problem will be to get heat out of the middle of a layer design. Also SC can allow for higher frequencies because of this. 

 

SC is currently used in high power transformers, the hope is it will make it to computer semi conductors 

or just switch to graphene too.

[Spindel]

6 minutes ago, Letgomyleghoe said:

or just switch to graphene too.

Graphene has been 5 years away from large scale production for the last 20 years. At least silicon carbide is a real thing, if SC is suitable for CPU applications is a different story. 

[Spindel]

33 minutes ago, Drama Lama said:

Yes but people don’t want phones to be hot

It's not necessarly the surface temperature that will be high, but if you start to stack transistors you will have problems to get heat away from the middle, and for that to actually work you need a semiconductor that can still operate at high temperatures unlike the "pure" silicon we use today in CPUs.  

[NelizMastr]

17 hours ago, JoostinOnline said:

Realistically, whatever marketing comes up with.  Switching units is always awkward, because marketing departments don't like how it affects numbers.  If people are used to the idea that a smaller number is better (even though that's not necessarily true), picometers won't go over well.

 

This is assuming there's no new kind of technology in place of course.

Back in the 90s the process node was still described in micrometers. While you could call the later Pentium III a 180nm design, it was commonly called .18um. 

We were basically in the same spot there. It will probably not be marketed in picometers until we've reached double digits. 

[Aereldor]

500 picometer?

[A51UK]

I am not sure where if is the point that CPU cannot getting small but we may end up with a different type of computer, here some type I have hear people been working on for year (e.g. DNA, Part Quantum, Pure Quantum , Artificial Neural Network,..) .  The Part Quantum is different from pure one as it cannot and does not use the superposition so it work more like how computer we use no, where the pure one allow for superposition.  We may end of with a mix of different type of computers system in one. 

[JoostinOnline]

3 hours ago, NelizMastr said:

Back in the 90s the process node was still described in micrometers. While you could call the later Pentium III a 180nm design, it was commonly called .18um. 

We were basically in the same spot there. It will probably not be marketed in picometers until we've reached double digits. 

I'm not old enough to remember that, but I figured that's what would happen given how HDD's were advertised when I was a teenager.  Terabyte wasn't widely adopted for awhile. They were 1000 Gigabytes.

[whm1974]

Carbon Nanotubes. if We can figure out how to make and purify large numbers of the Correct type, they could easily replace Silicon in many applications.

 

Althrough I'm sure they will be still expensive for Consumer Uses.

[sounds]

The biggest problem CPUs and GPUs have right now is how to get data moved around fast enough. Caches and faster memory help but if you look at actual memory latency (measured in nanoseconds, not bandwidth) memory is what's holding us back.

 

Shrinking the transistors does not change physics. The Apple Silicon M1 moves the RAM closer and that does change the physics. Well, shrinking the transistors can reduce the power usage. But heat isn't that big of a deal. Die stacking is not cheap, AMD and Apple are already using an interposer and that gives them the edge. Intel will likely start using chiplets and an interposer soon.

 

Maybe we'll get down to 3nm before quantum tunneling of electrons becomes so pronounced that the transistors can't switch anymore.

 

My bet is on CPUs that have memory built-in. You think 128 MB of "infinity cache" is big for the 6800 XT, but my bet is next-gen GPUs have close to 1 GB of "cache," and pretty soon, all the memory is "cache." Cache used to be too expensive - both in terms of transistors used and it generates lots of heat. I'm betting it doesn't stay "too expensive" for much longer.

[big oof 1]

easy.  what comes after 1 nanometer?  two nanometers, or 0 nanometers depending on which way youre counting

[Mark Kaine]

btw you can bookmark this, there won't be a "1nm" silicone based consumer CPU or whatever, at some point its not going to be worth it anymore (around 2-3nm) that was figured out many years ago already, Moores law is indeed dead and the industry will have to move on if they want to keep their growth and win margins (and they of course know that) doesn't mean they're aren't going to milk the silicone train dry before the inevitable move to [none silicone based chips] 

 

*bookmark it* quote me in 20 years