What happens when we reach 3nm? its the future more cores?

[Windows95]

We are getting close... 2030, lets say we get to 3nm, what then, just keep adding more and more cores?

 

How will a 3900x stand by then vs top of the line?

 

Let's compare 3900x with my current 10 year old computer for reference:

 

https://cpu.userbenchmark.com/Compare/Intel-Core2-Quad-Q6600-vs-AMD-Ryzen-9-3900X/1980vs4044

 

Can we expect this sort of difference?

[Geoff35674567]

none of them will run windows 95

[Animal901]

1 minute ago, Windows95 said:

We are getting close... 2030, lets say we get to 3nm, what then, just keep adding more and more cores?

 

How will a 3900x stand by then vs top of the line?

We have reached the point where cores are irrelevant. They'll  have to come up with 128bit to justify the continuing improvements. X58 still holds up to a point.

[Fasauceome]

We can do sub 3nm, might as well keep going until we have a new way to make CPUs

[leo1798]

We can definitely expect that sort of difference if not more now that there is actual innovation in the CPU space again. We will also certainly go below 3nm, likely before 2030 even. We (likely) won't see more than 32 cores for consumer desktop as gaming would not scale well beyond this point without ridiculous optimization, but workstation and server CPU's can likely go much higher as virtualization would actually allow the extra cores to be used.

 

P.S. Stop using Userbenchmark it's a terrible website with absolutely zero credibility. The difference between those two CPU's is even greater than the website claims because their weighting system is shit. 

[LogicWeasel]

14 minutes ago, Fasauceome said:

might as well keep going until we have a new way to make CPUs

There may be a future where they make CPUs out of something that isn't silicon to find more gains.  For reference, my dad worked for CRAY for a number of years (the Supercomputer company).  He noted how Seymour Cray had tried for a period to develop supercomputer chips out of Gallium Arsenide instead of silicon as it has certain advantages in performance and output.  Problem was the yields for a working wafer (even just a few viable chips) were so awful he had to scrap it.

 

You give a team of people with ideas like Seymour Cray (rest his soul, he was killed in a car accident far before his time) enough time and money, and they'll find something that isn't silicon to make CPUs on and make them better.

[Geoff35674567]

1 hour ago, LogicWeasel said:

something that isn't silicon

organics

[Kc7vwc]

Also quantum computing will come into play more in the next couple decades.

I personally look forward to entirely new architectures outside of x86, and new physical formats. No more wafer chips in sockets. 

Once upon a time, HP floated the idea to make CPUs as bricks, with interconnects all along the top & bottom. When you needed to upgrade, you bought another brick and just stacked it atop the previous one!

Ideas like those are what I look forward to.

[JoostinOnline]

2 hours ago, Fasauceome said:

We can do sub 3nm

 

2 hours ago, leo1798 said:

We will also certainly go below 3nm

Not necessarily. We're already at the point of the transistors being so close that the electrons interfere with each other.

[Geoff35674567]

18 minutes ago, VegetableStu said:

would decomposition be a thing? o_o

god was the OG in planned obsolescence

[Mister Woof]

10 hours ago, Geoff35674567 said:

organics

Bio neural gel packs from Voyager!

[Mira Yurizaki]

I would argue that shrinking transistors for a while was a way for many a manufacturer to employ designs that just meets requirements and not pursue anything better, because they can just clock the part faster or something.  If anything, there's still a lot of work on the software end as well because many a developer also relied on Moore's Law to make up for their deficiencies. So I think there's still room for improvement with what we have, but few are actually putting the work to realizing those improvements.

 

On the flipside though, I've seen companies looking into other materials for ICs. Like using gallium arsenide to make chips as it exhibits superior electrical properties over silicon.