Will humanity be able to shrink microprocessors below 12-9nm or we are condemned to reach the max computational limit in 6 years?
yes, of course
>>51425079
>Will humanity be able to shrink microprocessors below 12-9nm
Yes.
> or we are condemned to reach the max computational limit in 6 years?
There is no such thing, and area scaling is not the only way to improve performance.
>>51425079
Die shrinking is so last decade.
We die stacking and power efficiency now.
>>51425150
>what is quantum tunneling
>>51425568
Its something not even remotely related to this discussion.
>>51425776
quantum tunneling will be the key to future processor transistors, trust me mate i work at geek squad
3D processors and 3D ram yo
imagine....3D heatsinks though
>>51425824
lol.
graphiene, but first stacking just like memory
>>51426487
how far can graphiene go?
>>51427419
muh 400 ghz cpu
terahertz when
>>51427454
you weren't exaggerating when you said 400ghz
http://www.extremetech.com/extreme/175727-ibm-builds-graphene-chip-thats-10000-times-faster-using-standard-cmos-processes
>In theory, graphene is capable of operating at frequencies as high as 500GHz, well beyond any other material currently used in RF applications.
>>51427674
Yeah I said 400 precisely because I've read about it before
I also remember reading something about graphene "not being able to be turned off" or some shit like that
>>51425824
In 50 years maybe.
>>51427726
I wonder if single core CPUs with insane clockspeeds and IPC would be viable for a while once graphene process becomes a thing.
If human's computational power can fit in his head then we can achieve at least that.
R-right?
>>51427923
Yes.
Though it uses a completely different architecture for getting that sick energy efficient computing (synapses that are asynchronous, no constant clock to drain energy and synchronize things), which means it is good for some things and bad for some things.
I think the IBM neural chip project is targeting human-level neural computers with 1 kW power usage in"the future", or perhaps 15-30 years, with their SyNAPSE program that is developed with DARPA.
>>51428087
Organic computers utilizing modified human brain tissue. You have to feed it and dispose of its waste in a designated location.
Singularity soon, bros.
>>51425776
Quantum tunneling IS important to this discussion. Basically transistors are a switch. The switch is a piece of material that can be made conductive. The length of this material is usually what is referred to as the size of the architecture. It can only be made so small, because at a certain distance, electrons can tunnel through the switch, leaving you with something that is always on. So unless you re-design your transistor, you can only make current designs so small.
>>51425079
just make 2 physical CPU's that will double the speed!
>>51426487
Cell phones already do this, this is how we have 4Gb GDDR5 phones and tablets. Those run android.
>>51428204
>designated location
You did this on purpose didn't you
>>51425079
instead of making cpu's smaller and wafer like, why don't we make them taller like cubes, but not just cubes we could make them like dodecahedrons and shit
>>51425079
Lol I saw that thread on /sci/ too. That guy was retarded
We might be able to, but the cost of doing so could be prohibitively expensive.
>>51425079
>what are quantum computers
>>51426487
>>51429623
The purpose of die shrinking is not mainly to save space. It's to decrease lane sizes and thus resistance. When you stack your CPU, you don't decrease resistance, so you won't have any advantage apart from having a physically smaller sized die. If you stack twenty layers of chip on top of each other what you get is a chip that uses twenty times the power. Stacking makes sense for memory/storage where the limiting factor is space and not heat generation (at least not as much), but not really for CPUs.
>>51429623
Also, note that heat dissipation is almost impossible with a large three dimensional structure such as a cube. The core will just melt, as heat can't get out quickly enough.
>>51429717
Quantum computers are a completely different thing and won't help with traditional computing. Quantum computers will likely be way worse than traditional computers at most stuff we do. They will more likely just be co-processors in specialized super computers to solve problems that can be calculated more efficiently with quantum computers, such as factorization and deconvolution. At least for the next few decades. Maybe it will be used at some point for AI applications where way more efficient algorithms are possible as well. But it still won't be able to completely replace fully digital computers.
>>51429717
Quantum computers aren't necessarily faster than regular computers. Universal quantum computers might even be impossible.
>>51428910
>trying to describe a transistor to someone in EE
>utterly failing to do so
Transistors are a gate that sits upon the channel. What pop sci dipshits keep whining about is called the short channel effect. The reason why we're no longer using planar gates is to address the short channel effect, and we'll use topologies with progressively better isolation and control of the channel as we continue to pursue linear area scaling.
Try actually taking a look at the picture I posted.
>>51429689
EUV and 450mm wafers drive costs down. How Samsung fares with their laser produced plasma light source will have big implications for their future process nodes.
Utilizing less complex thin body FD-SOI nodes for lower power chips as well will see costs lower at smaller nodes.
7nm is already on the horizon, though quad patterning was utilized. If GloFo using IBM's resources really wanted to push hard enough they could bring that node to market within 2 years. And again if EUV pans out in a timely fashion then quad patterning wouldn't be necessary.
>>51426243
>imagine c3D heatsinks
b8
>>51429785
7 nm might be the limit for CMOS scaling. It might not be economical to go any smaller: http://m.theregister.co.uk/2013/08/27/moores_law_will_be_repealed_due_to_economics_not_physics/
For some reason I'm thinking about nano processors, but how would they maintain performance without overheating?
>>51429943
>7 nm might be the limit for CMOS scaling
Its not.
Articles like these are essentially nothing other than clickbait targeted at people who know nothing about the industry.
Watch an Applied Materials keynote sometime. The slide I posted at the top of the thread is from one.
>>51429954
What's a nanoprocessor?
