![1452998326200[1].jpg](https://i.imgur.com/9pTmpyF.jpg "1452998326200[1].jpg")
Moore's law for supercomputers appears to be slowing down, with progress more limited than had been anticipated. Have we hit the wall?
![supercomputers-and-superintelligence[1].jpg](https://i.imgur.com/7fOXoOb.jpg "supercomputers-and-superintelligence[1].jpg")
>>52464264
Sauce:
http://www.unz.com/akarlin/moores-law-for-supercomputers/
>>52464264
I hope not. We'll see in a few more years how it develops.
>>52465835
The professor of my high performance computing class told us it was indeed slowing down.
>>52467814
He said that we probably wouldn't reach exaflops within 2020
>>52467814
Silicon transistors really are slowing down.
If you look at technology over a longer term, it tends to plateau a little until a breakthrough is made then it shoots up again. Graphene transistors or quantum computing is coming in the next 100 years to revitalise moore's law. Which isn't really much of a law than an observation.
>>52467821
Graphene is only good for like 2 nodes. It's not going to change anything. Quantum computing isnt going to do shit for everyday people.
>>52464264
Silicon transistors are reaching their smallest size rapidly. New architectures will have to be employed to revitalize Moore's law. Otherwise, new materials will have to be used. Graphene looks to be the next small step before quantum computing really takes off.
>>52468141
Since this thread is about Moore's law for supercomputers, nobody gives a shit about everyday people, whoever these everyday people might be.
>>52468264
A quantum computer doesn't do classic computing so it wouldn't have flops.
>>52468141
"The average computer in 2000 will fit inside a house"
We've reached the point of good and cheap enough.
I don't really see how processors made of exotic materials will matter in the long term.
>>52468307
You can put a quantum computer in your house all you want. It's not going to do shit unless you happen to be solving extremely large size problems.
>>52464264
Regardless of what his happening, Kurzweil is hardly BTFO. In his language, a tapering of the curve merely signals the end of a technological paradigm and the beginning of a search for a new one.
>>52468339
What an insanely myopic view.
You're not interested in having the computational power of your current desktop PC fit on a chip the size of your thumb nail?
You're not interested in single nodes with the power to simulate the whole human brain, when today's best supercomputers must be content with covering only a small fraction of it?
You're not interested in nanobots that could mimic the function of red blood cells or neurons?
>>52468382
No because human were never ment to play God.
CPUs are old and busted GPUs and other ASICs are the new hotness
>>52464281
>Metabolome
>Proteome
Apparently computers will start to make up new words in 2045
>>52468420
An even more retarded concept.
So you're one of those people who took the flight of Icarus to heart and was traumatised by it...
>>52468466
https://en.wikipedia.org/wiki/Metabolome
You shouldn't flaunt your lack of knowledge like that, anon. What will your family think?
>>52467821
>quantum computing is coming in the next 100 years
by "next 100 years" you mean "last year", right?
http://www.heise.de/tp/artikel/46/46838/1.html
>>52468523
That's not a real thing tho.
>>52468270
>it wouldn't have flops
sure it would. It works be like every other computer, except algorithms can be in multiple states at once.
example: today a video-card has thousands of ALU to compute the same thing with just different input-values. Those thousands of ALUs could be in theory be replaced by a single quantum ALU.
>>52468523
I mean the point at which it beats conventional computing for the same price.
>>52468420
Luckily the concept of god is a silly caveman superstition so we can play what we want.
We should worry about not accidentally (or purposefully) destroying ourselves instead of playing fictional entities.
Best not open an umbrella indoors
>What an insanely myopic view.
>>52468382
Honestly I am more terrified of the implications of such technologies, we already see what humanity uses the internet for. Ordering dragon dildos and watching others play video games for them or just sending naked pictures to each other.
Also I strongly doubt that even the smartest minds will have the capacity to impliment these technologies if they actually become a reality. I mean software written today is still single threaded and the most brilliant people at sillicon valley are trying to reinvent another hype driven chat client.
I think the biggest breakthroughts in science are going to be within genetics and automation rather than IT.
>>52468733
>genetics and automation rather than IT
those breakthroughs will base on a breakthrough in IT.
>>52464264
Moore's Law was about price, not power you stupid nigger.
>>52469578
It was actually about transistor counts but nice try dude
>>52469584
https://www.youtube.com/watch?v=ylgk3HEyZ_g
You might want to argue that with Gordie.
>>52469188
Except the breakthroughts in IT are more or less over. Smart phones were the biggest recent development and also have hit a plateu.
>>52469584
https://youtu.be/hrtyUnad7Jo?t=618
Here's another one.
>>52469600
>>52469714
>youtube videos are an argument now
Thanks but I'm not an idiot.
>>52469747
It's Gordan Moore talking about his "law." Maybe you are an idiot.
>>52464264
It's probably simply intel doing nothing because amd is not competing.
>>52469747
11:01- "I wanted to get across "here's an idea, where the technology is going to evolve rapidly and it's going to have a major impact on the cost of electronics. That was the main point I was trying to get across, that this was going to be the path to low cost electronics."
- Gordan Moore.
>>52464281
>>52464264
>conflating transistor count with performance
Completely disregarded.
>>52467821
>Silicon transistors really are slowing down.
Performance is derrived from architecture first and foremost, the issue is not the substrate material. This is popsci garbage regurgitated by the clueless.
No one is going to massively increase serial integer performance without some radical paradigm shift in IC design. We can't possibly feed instructions to the ALUs fast enough. FPU performance can still make healthy gains by way of larger vector instructions, and multicore scaling can still benefit greatly without any major R&D being required, especially when it comes to socket to socket communication. The structure of the core itself is the issue, and this is true regardless of ISA. Every Von Neumann architecture will reach a plateau in single core performance.
>>52468307
"Within 50 years robots with human intelligence will replace humans"
