Moore's Law: the end is near-ish!

Moore's Law: the end is near-ish!

Summary: Many agree that one of the key driving forces of the computer revolution is the ability to provide consumers with devices of ever increasing power. Every year manufacturers put out a new line of more powerful products – twice as powerful, in fact, every 18 months. And, if we can believe Michio Kaku, in his book the Physics of the Future, this is about to come to an end.

SHARE:

Many agree that one of the key driving forces of the computer revolution is the ability to provide consumers with devices of ever increasing power. Every year manufacturers put out a new line of more powerful products – twice as powerful, in fact, every 18 months. And, if we can believe Michio Kaku, in his book the Physics of the Future, this is about to come to an end.

Moore’s Law is named after Intel co-founder Gordon E. Moore, who described the trend in his 1965 paper Cramming more components onto integrated circuits.

“The future of integrated electronics is the future of electronics itself. The advantages of integration will bring about a proliferation of electronics, pushing this science into many new areas.”

His predictions were uncanny. Predicting the role in integrated circuits in homes, personal computers, automobiles, and what he called personal portable communication equipment. Moore’s Law has been the keystone in long-term planning in the semiconductor industry for research and development. In his book the Physics of the Future, Kaku suggests that this is about to come to an end.

The laws of physics are to blame. The current process allows us to create chips with transistors that are measured in atoms. Eventually, or about 2020, transistors will become so small that quantum theory or atomic physics will take over and electrons will begin to leak out of the wires.

In 2020 it is estimated that we could get to the point where we have the ability to create transistors that are 5 atoms wide. According to Kaku, at this juncture the Heisenberg uncertainty principle comes into play, which states that you cannot know both the position and velocity of any particle, which means we cannot know where an electron is, and therefore cannot be confined to a wire, causing a circuit to short circuit.

Does the end of Moore’s Law equate to the end of Silicon Valley?

The bad news is that we have about 8 years before this specific etch-a-sketch method of chip development will slow to a crawl. While innovation will likely continue, the time we will have to wait longer for our computer power to double.

The good news is that we are now entering a new paradigm where quantum mechanics promises a similar technological disruption, according to Science Daily's recent article Single-Atom Transistor Is End of Moore's Law; May Be Beginning of Quantum Computing. Researchers are working to build a single-atom transistor as a first step to the development of a quantum computer that works by controlling the electrons and quantum information, or qubits.

But don’t change the name to Quantum Valley anytime soon as this technology is a ways away. Earlier this year at Intel Investor Day, Mark Bohr who overseas much of Intel’s processor research, was quoted as saying, The end of Moore’s Law is always 10 years away. And, yes, it’s still 10 years away.

While I am certain Intel is innovating in attempts to move passed the barriers of atomic physics, I hope that they are not taking a position similar to the position RIM took as a response to the touch screen (i.e., they created a device where the entirety of the screen needed to be depressed to make a selection!).

What do you think, is the end near for Moore’s Law? Let me know.

Topics: Processors, Cloud, Hardware, Intel, Smartphones, PCs

Gery Menegaz

About Gery Menegaz

Gery Menegaz is a Chief Architect for IBM with more than 20 years supporting technologies in the financial, medical, pharmaceutical, insurance, legal and education sectors. My Full-Time Employer is IBM. I write as a freelancer for ZDNet.

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.

Talkback

47 comments
Log in or register to join the discussion
  • eventually yes it will end...

    But I can find articles from 1995 in a google search saying the 250nm k6 was pushing moores law to the breaking point..
    TrishaDishaWarEagle
    • 250nm was a barrier

      The 250nm was a purely technological barrier this however is a physical limit of the silicon medium, there are however a lot of promising alternatives to take us past silicon and which one will finally succeed silicon I don't know, non the transition will no doubt serve as a major mile stone in the history books.
      KevinMaguire4
    • colours surely? or light?

      Hi :)
      Perhaps binary/digital computing is coming to an end. Moving to photons carrying data instead of electrons. If colour plays a factor rather than presence/absence of a photon then it's heading towards analogue computing but not quite the same as last time.
      Regards from
      Tom :)
      Tom6
    • AMD says end of Moore's Law is near

      This time it's happening, and AMD agrees, too. You simply can't go smaller than 1 atom, and at that point you need to deal with quantum effects anyway, so we'll need quantum computers.
      gremmor
  • No, a loophole in Moores law will always be found.

    Soon we will be dealing with three dimensional curcuits, 50 core CPUs, then faster ram, larger cache sizes faster drives. The moment that the speed increases stop is the moment that most people lose their primary reason for purchasing new computers and no new computers is bad for intel's bottom line so they will find some way to continue to add increases in speed or they will probably collapse.
    balsover
    • I agree, partially

      Moore's law states that the electronics will double in power and half in size. We are only nearing the limit of the second part of the equation. Increasing power will continue, possibly at an even faster pace than before, since the technology will have reached it's limit as far as shrinking the equipment goes. When increased power is the only concern, it will be quite easy to add more circuits.
      purevw@...
      • But while this is true

        Only the size of the CPU's really will be shunk to the max, then everything else needs to be shrunk aswell, and while it can be done using the same technology, I feel as that will be something that is done.
        Jimster480
    • But

      Intel makes chips, not computers. Perhaps they'll go back and and figure out how to rework their earlier chips so that they can run faster.

      Gods know that I'd love to switch out the old dual core 1.25 GHz in my old G4!
      RangerJimK
    • Many Reasons for 'The End' of Moors Law

      There are many reasons for expecting the end of the Moors Law period of electronics advancement.

      First, limits on size exist. We have know since 1992 that the smallest possible transistor takes around 7 atoms. IBM did this with electron force microscopes. Current transistors in State of the Art chips use around 100,000. Trace wires don't appear to work well if they are less than around 60 atoms wide. There have been some experiments that use benzene molecules to get a wire that is only around 5 atoms wide, but these only work at low temperatures.

      Second, limits on energy exist. We saw the raw speed of processors peak at around 2 GHZ for passive cooled processors, and at around 3GHZ for fan cooled processors around 5 years ago. Your processor today isn't really much faster. Oil or water cooling can allow a push up to around 7 GHZ, but no higher. The reason is that the energy dissipated in a chip is a function of the frequency. At these limiting speeds, we are dangerously near to melting the silicon chips.

      Over the past 5 years, we have not increased the power of the processors, we have just piled more processors on the chip. That is the real reason that parallel architectures are so important today. That is also why Intel and its' competitors don't advertise the clock speed so much any more.

      The promise of diamond and graphene is that the clock speed can be increased again. Silicon melts at several hundred degrees C, but Diamond melts or sublimes at temperatures of over 2,000 degrees C. Graphene is of course, a 2 dimensional version of diamond. but, even if we do successfully make the transition to diamond, that will only buy us a few more iterations until we reach the wall. Going into three dimensions only buys us a little gain, perhaps another generation. each generation is a year and a half.

      The limits of physics are well established. They can't be broken. We can design our way past several of the bottlenecks that are in our way, such as the current limits of Silicon, which we have reached, but the final walls are still in front of us.

      We still haven't made a single computer using quantum bits, so we don't know what the limits of that are. It was supposed to be magical, but now, maybe not so much. The current quantum wells require more atoms for construction than do the smallest transistors that have been made.

      2020 still looks like the end for Moors Law. It's been the expected end since 1995.

      Already, the manufacturers are making the dies bigger, rather than making most of the components smaller. But, bigger chips means slower chips, as the speed of light is still a limiting factor. Signs of the end are all around us. we are now at 20 nanometers for the smallest size components. New factories for each jump in size cost around 10X more than the generation before. building 45 NM factories cost around 2 Billion. The new 20 NM factories cost around 10 Billion each. We can expect that the jump to around 8 NM will cost close to 80 Billion. Then, there is the transition from Silicon to Carbon. That will take as much effort as did the jump from Gallium Arsinide to Silicon that was made in the early 1970's and late 1960's. That took 10 years. That may be why Moors Law has been slowing in the past 5 to 7 years.
      YetAnotherBob
      • We were doing so good, at least up until...

        "We still haven't made a single computer using quantum bits, so we don't know what the limits of that are. "

        Just wanted to correct you, since most people might not know this and take your word about our seeming lack of advancement/development in Quantum Computing:

        "10/31/2011 @ 12:00PM

        Lockheed Martin Installs Quantum Computer

        On Friday, Lockheed Martin installed its D-Wave Quantum Computer
        at USC’s Information Sciences Institute in Marina Del Rey, California.

        The D-Wave computer operates as an adiabatic quantum computer,
        and consists of a superconducting 128-qubit chip in a cryogenics
        system within a 10 square meter room. "
        [source: http://www.forbes.com/sites/alexknapp/2011/10/31/lockheed-martin-installs-quantum-computer/ ]

        So yea, while it's not a desktop computer by any means, it IS a Quantum computer existing outside of a lab, and more importantly it is available to purchase.

        And that concludes this evening's broadcast >insert TV Calibration Image and long annoying bbooooooooppppp
        Formula350
    • I disagree

      I bought a quad core 2.4 GHz Q6600 January 2008. It was not the top of the line. It benchmarks on PassMark at about 3900. There were faster 3.0 GHz systems up over 4000 at that time. It should have doubled three times by now(every 1.5 years). That would make a processor benchmark at 32000 by today. They are only at 16000 for the best Xeon's. So Moore's law has already slipped to at least 2.25yrs as far as power. I was also not impressed with this processor in one aspect. It's responsiveness was not an improvement. Coming from a Pentium 4 2.8 GHz CPU I saw little change and probably for the worse. If you look at these modern processors mine is clocking along at 1.6GHz until you hit a key on the keyboard.
      And as a computer programmer I know that it is not easy or possible to make programs run in parallel to take advantage of the multiple cores. And even when you can, you do not get a 100% increase per CPU. You are lucky if you get 60%. And cores are not doubling every 1.5 years.
      david@...
      • you are right

        you are right. try SSD hard drives , much of the speed problem you are experiencing comes from the hard drive not the CPU.
        docesam
      • Moore's Law isn't concerned with clock rates

        Moore's Law isn't concerned with clock rates. It was about two factors: the number of transistors in a given amount of space and the cost of those transistors. The prediction was that we would have a lot more and they'd cost less. That has been entirely consistent thus far.

        What speed those transistors operated at or how well we could exploit them collectively are separate issues that down come under Moore's law.
        epobirs
      • You are correct

        I remember in the early days of computers when you couldn't wait to buy the latest and how much faster they were than one just a year older... not the case anymore. Sure the latest and greatest is faster but nothing like the changes you used to see, most often the increase in speed is barely noticeable.

        So if you look at Moore's law as the end result we are getting in productivity with computers, then it has been been broken for a while now. Sure I know Moore's law is more about size and cost, but what does that matter if it doesn't translate into faster machines in the end? Sure we will be able to have smaller smartphones, but when it comes to desktop computers we are long past the days when the latest computer blows away last years model.
        RAK5
  • First we'll have intels 3d transitors, then other mediums like silicene

    with graphene connections. We may stop shrinking transistors but be able to stack and pack without heat build up and have 1024 core smartphone cpus. Well have memristors and optical lightpeak connections instead of the useless copper baby step apple took. The overall perf of the systems will continue marching forward via other breakthroughs regardless of moores law. Perhaps we'll have ubiquitous high speed cloud access to multi billion core datacenters and not need to rely on local cpu tech advances.
    Johnny Vegas
    • Heat is Your Enemy

      Heat is the limiter in modern MPU's. going 3D just gains you a little bit of extra cooling. If you try to go in real 3d, the exra heat generated will quickly overwhelm your design. That's why clock speed stopped growing in leaps and bounds.

      Now, there are 2, 4 or 8 processors on a single chip. But, multiple processor operation in operating systems always runs at degraded performance levels. Windows running on 8 processors is typically only running at about 40% of the throughput per processor as Windows running on one processor. Granted, 8 X 0.4 is still 3.2 X as much getting done, but the whole thing falls apart as you increase processors. That's why supercomputers don't run Windows.

      The whole limits thing is physics, not programming. Sorry, but we are very close to the wall. We've already hit the low walls a couple of times, and are against a couple more low walls. But the real high wall is within sight. Diamond based IC's with each atom of the circuit placed at precise locations is as good as it can get. That's in the labs right now.

      But, before you despair, remember that the real limits are still a couple of orders of magnitude faster than we have today, and about a million times denser.

      But, that is as far as it can go.
      YetAnotherBob
      • Yes but with graphene and silicene and such new materials there is much

        lower resistence, even at room temperature. So I think well be able to generate much less heat and therefore pack more densely not just because of shrinking transistor size, but because of not having the thermal issues that prevent stacking several layers of cores on top of each other in a single soc package. Also your windows processor numbers are wildly out of date. And not just for windows HPC, which btw supercomputers do run, but for typical datacenter, desktop, laptop skus as well. And soon WP8. Windows is now also sporting start of the art power management to dynamically disable as many cores as possible in those non HPC skus.
        Johnny Vegas
    • single > multiple !!!

      single core 3.0 Ghz CPU in most cases is much better that quad 2.0 Ghz CPU. 1024 cores will not do you any good in most (but not all) cases .
      docesam
  • Being hearing about it since 1980's

    Still waiting for the mystical ceiling ....
    wackoae
  • Wasn't Intel doing a lot of research on stackable transistors?

    That'd keep Moore's law going for quite some time.
    Aerowind