• Log in with Facebook Log in with Twitter Log In with Google      Sign In    
  • Create Account
  LongeCity
              Advocacy & Research for Unlimited Lifespans


Adverts help to support the work of this non-profit organisation. To go ad-free join as a Member.


Photo

Is Moore's law going to hit a wall soon ?


  • Please log in to reply
21 replies to this topic

#1 Oren K

  • Guest
  • 8 posts
  • 0
  • Location:Earth

Posted 03 June 2011 - 06:26 PM


Hello all,

I hear many people saying that Moore's law is going to hit a wall very soon... is it true?

I just heard that Intel has announced new 3D technology, that's very nice, but in order to continue the doubling in computers power every year for the next 30 years I think that we need much more than that...

So what are the experts saying about that? what new technologies are been developed for keeping the trend ?

By the way, for a nice video about this search in YouTube for:

Moore's Law Got Me!

Go to the first result, the other is blurry.

(as a new user I can't send links)


Thanks!

Edited by Oren K, 03 June 2011 - 06:31 PM.


#2 mikeinnaples

  • Guest
  • 1,907 posts
  • 296
  • Location:Florida

Posted 06 June 2011 - 01:14 PM

Soon? No... it is not going to hit a wall soon, though I guess that depends on what your definition of 'soon' is. I think the fledgling 3D technology Intel is releasing in its new processors will help bridge the gap to a complete paradigm shift, perhaps to something like quantum computing. Even without a paradigm shift to a completely new technology, I am thinking the 3D processors can expand Moore's law out easily past another decade. In technology, a decade of time is practically forever... so yeah.

sponsored ad

  • Advert

#3 Anewlife

  • Guest
  • 210 posts
  • 10
  • Location:Australia

Posted 27 December 2012 - 05:59 AM

The jump in processing power lately from each new architecture has diminished recently.
I was reading up on this at the beginning of the year and if I remember correctly its going to hit a wall in about 8-10 years, then we will probably go on to parallel processing as microsoft has suggested, multiple CPUs for each computer until newer technology is developed.

The new tech is very far away though, as in decades. I forget what it was called but we pretty much need another Einstein to do a physics breakthrough 1st.

Edited by Anewlife, 27 December 2012 - 06:00 AM.


#4 Mind

  • Life Member, Director, Moderator, Treasurer
  • 18,997 posts
  • 2,000
  • Location:Wausau, WI

Posted 27 December 2012 - 07:09 PM

I don't think we will hit a wall anytime soon, but the advances in computing power vs. cost (or energy input) will become more of a blend bbetween software advances, bandwidth expansion, AND processor design/feature size (not just from processor advances/feature size alone).

#5 platypus

  • Guest
  • 2,386 posts
  • 240
  • Location:Italy

Posted 27 December 2012 - 10:55 PM

I don't think we will hit a wall anytime soon, but the advances in computing power vs. cost (or energy input) will become more of a blend bbetween software advances, bandwidth expansion, AND processor design/feature size (not just from processor advances/feature size alone).

How will the bandwidth-expansion heppen, i.e. what are the technologies in the horizon?

My understanding has been that while computing power is growing exponentially (Moore's law), bandwidth grows a lot slower since there are physical limits on how much data can be sent inside an optical fiber?

Some near-future satellites will produce a ~Petabyte of data per year per satellite. Now, if everyone who is interested in the data is trying to download these data there will be some problems with network capacity...

#6 Anewlife

  • Guest
  • 210 posts
  • 10
  • Location:Australia

Posted 28 December 2012 - 12:14 AM

Bandwidth is not the problem at all, it is ample.

Advances in software (Java revisions, DX versions etc) and bandwidth expansion (cache size on the CPU, and RAM speed) dont compare to the performance increase of a new architecture.
If you built 2 computers, one with underclocked ram, older windows and an older JSE but the latest CPU architecture, and then one with very expensive overclocked RAM, latest updates, but the previous architecture, the 1st computer will be faster and possibly also cheaper as RAM tends to get expensive.

I have been benchmarking and overclocking for years and shits catching up, its already evident CPUs have had less and less headroom for overclocking, requiring people to resort to better cooling methods, and worrying about damage more than ever, Where as previously you could get a massive performance increase. That is a heat issue though, the real issue is the physics involved of the die size, something to do with the electrical signals being too big for the die.

A parallel processing example is SLI/Xfire. Eventually we will be buying computers with 2, 3 or 4 CPUs.


Look up quantum computer its the hopeful new tech thats still far away.

Edited by Anewlife, 28 December 2012 - 12:19 AM.

  • dislike x 1
  • Informative x 1

#7 PWAIN

  • Guest
  • 1,288 posts
  • 241
  • Location:Melbourne

Posted 28 December 2012 - 05:27 AM

Fusion has been 40 years away for many decades now. Same with limits to Moores law - been just a few years from ending for decades now. According to Kurzweil, it is not going to happen - we'll just keep making new breakthroughs. I think at worst we will see a paradigm shift or a change in underlying technology. I read somewhere that Intel has a whole new technology it is preparing and will release in a few years and that they expect a jump in performance at that time. Supposed to have some working stuff in relation to that too. We'll see...

#8 Mind

  • Life Member, Director, Moderator, Treasurer
  • 18,997 posts
  • 2,000
  • Location:Wausau, WI

Posted 28 December 2012 - 10:29 PM

Some possible bandwidth increase here: nanophotonics, seems about ready to move into the fab.

#9 Mind

  • Life Member, Director, Moderator, Treasurer
  • 18,997 posts
  • 2,000
  • Location:Wausau, WI

Posted 04 April 2013 - 05:57 PM

Faster switching memory enhances the overall performance of computing systems as well.

#10 knite

  • Guest
  • 296 posts
  • 0
  • Location:Los Angeles, California

Posted 28 July 2013 - 10:04 AM

Well, we are reaching the lower limits of silicon. Feynman once said, "There's plenty of room at the bottom", but we are coming up on the bottom pretty quickly (I think we are down to something like 14nm, with upcoming cpu architectures?).

It's going to require a fairly major shift to continue increasing processing power, but it will certainly be possible. The 2 huge advancements I see in the future are a shift away from silicon and toward carbon, and the possibility of room temperature super conductors. These are both very sci-fi at the moment though.

The next (somewhat realistic) major change I see in computing is memristors. Currently memory -> cpu is the big bottleneck in most computing.

Edited by knite, 28 July 2013 - 10:07 AM.


#11 ihatesnow

  • Guest
  • 776 posts
  • 251
  • Location:rochester new york

Posted 02 September 2013 - 10:39 PM

more about computers............... http://www.cnn.com/2...computerchips4p

Edited by ihatesnow, 02 September 2013 - 10:43 PM.


#12 forever freedom

  • Guest
  • 2,362 posts
  • 67

Posted 04 September 2013 - 06:15 PM

I'm very sorry to admit that CPUs have already hit a wall several years ago, concerning clock rate. We get several parallel CPUs nowadays but the performance increase is clearly not the same as before.


GPUs (graphics processing units), on the other hand, have been advancing steadily, to such an extent that latest supercomputers have been increasingly relying on GPUs and less on CPUs.

By the way, we may also measure Moore's Law effectiveness by the rate of increase of processing power of our greatest supercomputers, which have fortunately been advancing at a steady rate. We should have an exaflop supercomputer by 2020, at most, if not sooner, around 2018. Sorry guys i'm too short of time right now to search for a graph for supecomputer speed increase over the last decades.

As long as Moore's Law or supercomputer speed advance continues, until 2050 we will certainly have some form of strong AI.

#13 mehisa

  • Guest
  • 12 posts
  • 2
  • Location:u.s
  • NO

Posted 10 September 2013 - 07:18 AM

also computer programmers are using faster computers an excuse to make bloatware and inefficient programs. Really a 1995 computer was just as fast and simpler software.
  • like x 2
  • dislike x 1

#14 Layberinthius

  • Guest
  • 298 posts
  • 26
  • Location:Cyberspace

Posted 10 November 2013 - 05:56 AM

also computer programmers are using faster computers an excuse to make bloatware and inefficient programs. Really a 1995 computer was just as fast and simpler software.


I just had to say that I couldnt agree more with this statement.

I thought I was the only person on this planet with this viewpoint.
  • like x 1

#15 John Gritsman

  • Guest
  • 3 posts
  • 2
  • Location:Brazil
  • NO

Posted 14 December 2013 - 08:24 PM

The limit of silicon it's already reached in lab. Australian scientists made a one-single atom transistor last year.

http://newsroom.unsw...istor-“perfect”

#16 Mind

  • Life Member, Director, Moderator, Treasurer
  • 18,997 posts
  • 2,000
  • Location:Wausau, WI

Posted 15 December 2013 - 05:50 PM

Clock speed leveled off a few years ago. Not much chance of an increase until more photonics is integrated into chips.

#17 ihatesnow

  • Guest
  • 776 posts
  • 251
  • Location:rochester new york

Posted 03 May 2015 - 07:38 PM

http://www.scienceal...ree-atoms-thick

 

 

 


Edited by ihatesnow, 03 May 2015 - 07:43 PM.


#18 Avatar of Horus

  • Guest
  • 241 posts
  • 291
  • Location:Hungary

Posted 04 May 2015 - 07:43 PM

...
By the way, we may also measure Moore's Law effectiveness by the rate of increase of processing power of our greatest supercomputers, which have fortunately been advancing at a steady rate. We should have an exaflop supercomputer by 2020, at most, if not sooner, around 2018. Sorry guys i'm too short of time right now to search for a graph for supecomputer speed increase over the last decades.

...

 

Here is one, from the: http://top500.org/st...tics/perfdevel/

 

Exponential growth of supercomputing power as recorded by the TOP500 list

 

supercomputers_top500org.png

 


#19 treonsverdery

  • Guest
  • 1,312 posts
  • 161
  • Location:where I am at

Posted 18 September 2015 - 09:56 PM

Readers here might like reading about rogers materials lab http://rogers.matse.illinois.edu/

among the products are 40 nm thick ntergrated circuits

 

It is possible a plurality of 40 nm thick CPUs could be laser cooled from the base then placed at a staggered stack

 

_

 _

  _

   _

    _

 

that is a way large quantities of integrated circuits could function with laser cooling at a particular volume

 



#20 JFoulds

  • Guest
  • 3 posts
  • 2
  • Location:United Kingdom

Posted 25 December 2015 - 10:54 PM

Moore's Law, which is the observation that the number of transistors per square inch of physical space roughly doubles each year, came to an end a while ago.

 

Parallelism, however, is still an ongoing field. This being said, we're now starting to hit physical limits, predominantly due to the physical space on the processors vs heat dissipation for the physical devices. On-chip photonics (which would allow further apart physical components to communicate as if they were essentially adjacent) is an interesting developing area in this field, and there are a couple of other potential advancements which would allow 'performance' to increase in a Moore-similar pattern in future years.



#21 BadApples

  • Guest
  • 9 posts
  • 2
  • Location:Boom Belgium

Posted 27 December 2015 - 11:13 PM

We are at 10 nm, there is a proof of concept in which bit flipping occurs on memory chips (that can be induced with software) as electrons travel more distance then initailly thought.

 

Not much of an issue, resources can be used in parallel and don't have to be in the same physical location. 

 

But realtime computing won't be happening anytime soon at the current pace.

 

BA

 



sponsored ad

  • Advert

#22 dihard

  • Guest
  • 2 posts
  • 2
  • Location:USA

Posted 16 October 2017 - 10:59 PM

The official Moore's Law is regarding how many transisters you can fit into the same space. However a lot of people tend to use it in a looser sense of general processing power which has other factors like clock rate and efficiency gains like hyperthreading, parallel cores, etc. Depending on how you are using it you'd hit a wall at different points. 






1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users