At some point the semiconductor manufacturing process is going to be so small that it's not going to be able to be shrunk anymore. Probably in the next 10 years. What will happen to the state of cpu's and gpu's? Will we just add more cores? Will the ever increasing power of computers start to slow dramatically?
What happens when semiconductor process cannot be "shrunk" any smaller?
- Thread starter cobra1981
- Start date
That's a very good question. I believe Intel has a plan for an 8 nm process, but with the effects of quantum tunneling it's going to be pretty difficult to go beyond that using silicon substrates. In the long term, there will probably be some effort to shift to a different form of circuit manufacturing that can be shrunk smaller.
I believe there is has been some research into making logic circuits from carbon nanotubules, and potentially even DNA. There aren't really any definitive timeframes for bringing such technologies to market though.
Before the industry shifts away from silicon, there will probably be a tricks they can try. There will probably be a focus on making new more efficient architechtures. We may also see a trend toward larger dies, as yields improve on the smallest economically achievable silicon process. Beyond that, we may see use of die stacking to add more transistors.
One thing is for certain, the semiconductor industry will innovate to improve performance. If there aren't material performance increases between generations of products, people and corporations won't feel the need to upgrade, and that upgrade itch is what drives the industry.
dennilfloss
Past Lifer 1957-2014 In Memoriam
The software engineers will have to relearn how to optimize code and be less sloppy, just like in the early years.
Actually, improvements in algorithms already outstrip improvements in processor power. If there's money to be made by doing something faster, you can bet that people are making the software run faster. And as others have already pointed out, moving off of silicon will allow for better processors.
The scaling will continue but it will come in terms of materials scaling (see this post) for performance and vertical scaling (chip stacking) for density.
The path down to 4-5nm is already seen as feasible/doable. Its the post-4nm stuff that has everybody in R&D scratching their heads right now.
Considering that we're on 32nm right now, 4-5nm is a long, long way off. By the time we actually get there, there's going to be another road map stretching off into the future and we're going to be wondering how they're ever going to solve the problems at the next end point.
Last edited:
Perhaps there will come a day when we can't rely on hardware to automatically get faster every year. And then, the improvements will be in the customizations available, opening up the market to small-scale entrepreneurs and innovators
http://www.bunniestudios.com/blog/?p=1863
This blogger talks longingly of the return of artisan engineering, and the rise of such things as heirloom laptops. Heady stuff!
http://www.bunniestudios.com/blog/?p=1863
This blogger talks longingly of the return of artisan engineering, and the rise of such things as heirloom laptops. Heady stuff!
I read the OP's question as when will Moore's Law break. I agree with the prior comments that we can figure out how to make it work in the future even if we do not know how today. However, that does not mean that it will not break. Moore's Law works because every time they cut the price per transistor, people buy more transistors. Moore's Law is a response to economic demand. When the demand for transistors is sated, Moore's Law breaks.
Blitzvogel
Platinum Member
- Oct 17, 2010
- 2,012
- 23
- 81
there's money to me made in making slower, if that's what you mean.
In the old days it was called the Wintel alliance. People with slow computer buy new one, which is a new Windows software license. Both Intel and Microsoft win.
I believe it ll go as:
efficientcy goes up, after that/during, you start stacking things.
When that doesnt work anymore, something new comes along, like say Graphene/optical (I think we might "skip" Graphene, if research in optical electronics is advanced enough by then (depends on how long scilicon can be pushed)).
All in all, we *could* be looking at silicon for another 25years or something (thanks to process nude shrinks, makeing more effecient designs, and stacking), so its not like its right around the corner.
efficientcy goes up, after that/during, you start stacking things.
When that doesnt work anymore, something new comes along, like say Graphene/optical (I think we might "skip" Graphene, if research in optical electronics is advanced enough by then (depends on how long scilicon can be pushed)).
All in all, we *could* be looking at silicon for another 25years or something (thanks to process nude shrinks, makeing more effecient designs, and stacking), so its not like its right around the corner.
sm625
Diamond Member
- May 6, 2011
- 8,172
- 137
- 106
We're at 22nm now.
There are a number of full nodes between now and then, but not a massive amount.
Intel process roadmap showing the next 10 years. In 10 years (give or take) we will be at that 4nm figure if everything goes to schedule.
Which has what to do with major companies developing software for their own internal use? If Google can speed up their search algorithm by tweaking the code, they have no reason to keep their old, slow code. It just costs them money.
Great, so where can I buy a 22nm chip?
It would appear that 22nm isn't quite production ready yet or that Intel is holding it back to further improve the yield. AMD is going to be on 32nm for the foreseeable future, and ARM chips and GPUs will only hit 28nm later this fall. There might be a few places making RAM or SSDs that are using something closer to 22nm, but by and large we're not quite there yet.
Cogman
Lifer
- Sep 19, 2000
- 10,284
- 138
- 106
What will happen?
My bet is that the focus will be placed on new and different hardware components. More than likely, we will see the hardware do more and more of today's software jobs. For example, an inbuilt MP3 decoder, physics engine, or whatever. Either way, hardware will likely do a lot more work than it is currently doing.
Other changes we might see involve the very design process. We may see more emphasis in async circuits (which consumes less power and are faster than synchronous circuits, but are significantly harder to design).
We might start to see more different types of components being commonly used in CPUs. Right now, CPU design is pretty much completely dependent on the transistor. We might see more exotic components or even transistors being invented.
My bet is that the focus will be placed on new and different hardware components. More than likely, we will see the hardware do more and more of today's software jobs. For example, an inbuilt MP3 decoder, physics engine, or whatever. Either way, hardware will likely do a lot more work than it is currently doing.
Other changes we might see involve the very design process. We may see more emphasis in async circuits (which consumes less power and are faster than synchronous circuits, but are significantly harder to design).
We might start to see more different types of components being commonly used in CPUs. Right now, CPU design is pretty much completely dependent on the transistor. We might see more exotic components or even transistors being invented.
To some extent this is already happening.
The average consumer does not really need faster anymore. Apple sells on design rather than speed, as do netbooks and tablets.
We're at 22nm now.
There are a number of full nodes between now and then, but not a massive amount.
Intel process roadmap showing the next 10 years. In 10 years (give or take) we will be at that 4nm figure if everything goes to schedule.
I will be EXTREMELY surprised if the cadence stays at 2 years. There is less need than ever to push the pace, This roadmap will slip dramatically, I bet.
Phynaz
Lifer
- Mar 13, 2006
- 10,140
- 819
- 126
That's not the point.
The fact is, Intel today is manufacturing chips on a 22nm production process. From a technology point, he is correct, we are at 22nm today.
But they're not producing them mass-market yet, so either the process isn't ready or it's not cost-efficient yet. Intel can also most likely produce chips on the 16nm process they're developing, but you surely wouldn't claim that we're at 16nm. We're almost at 22nm and in a few months when Intel starts their production run, I'll agree that we're there. If they've already started that final production run, then I'd say we're there. As far as I know they have announced such an event so it would come down to speculation.
Either way the argument has digressed to semantics and probably isn't worth carrying on at this point.
Intel has to, otherwise they stand to lose revenue.
So long as the revenue growth exists from people throwing away their computers from two years to replace with one today, the money will be there to keep making computer tech that compels you to throw away that otherwise perfectly working computer.
Edrick
Golden Member
- Feb 18, 2010
- 1,939
- 230
- 106
Agreed. Just because you can not buy one yet does not mean the technology is not alive and well (and in testing).
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter