Friday, November 26, 2010

Moore's Law

In the Computer world, there is a term called Moore’s Law. It was pioneered by one of the founders of Intel, the leading semi-conductor corporation worldwide. Essentially this means that processor speeds will double every two years due to more transistors on silicon. Nvidia’s GTX 580, a graphics card released last week, has 3 billion transistors. The Pentium Processor, the original one, had 2 million transistors. This was released in 1995. That is 1500% increase in the number of transistors in a 15 year span. The chips have also been getting smaller down to 32 nm today. This is meant to reduce power consumption and increase efficiency. However, due to the limitations on small a chip can be, Moore’s Law may becoming to an end. Sandy Bridge, Intel’s new microarchitecture, will be released on the 32 nanometer technology and its successor will be manufactured on 22 nm technology. However, by 2020, the chips will get to a point where they are too small. 

Intel’s main competitor, Advanced Micro Devices, is attempting to attack the Netbook market that was powered by Intel’s Atom Processors, with an APU (Accelerated Processing Unit). This is the combination of a processor and a graphics processor, the brains and eyes of a computer. This revolutionary product is set to be more efficient, and strike Intel across the bow. AMD has been losing money had it not been for the Anti-Trust violations of Intel that were paid to AMD. Intel has a wide advantage at the moment, and Intel is expected to maintain its performance advantage with Sandy Bridge. It is expected to bring a 25% boosts while using 10% percent less power. But one question on the minds of many in the industry is will there be this exponential growth in performance in the Future? Time will tell. This greatly effects other industry and the American semiconductor industry. 
-S. Martin 

4 comments:

  1. You omitted the part of Moore's Law of Technology that states that both the physical size and monetary cost of technology will halve annually. That is the main component of the Law, and therefore your claim that the Law will come to an end because the size of the chips is decreasing is flawed because this means that the main component of the Law is in fact occurring.

    -T. Taylor

    ReplyDelete
  2. This is true but the point is that the chips can only get so small. Thus performance will not be able to come as they did other Moore's Law. The advances will be either more expensive or happen slowly. thus moores law as we know it will either be altered or invalidated.

    ReplyDelete
  3. Since Moore's Law will not reach $0, nor will the area of the chips be too small to build - Moore's has finite application. For a complete understanding of AMD's cognitive leap return to those days of yesteryear: Look at the original logic cycle of vonNeuman and Eckert and today's Buss processing cycle. The future will always be doing 'something' differently. We are now in parallel processing cycles with the unitary chip. We will expand the I/O buss to integrate it with the processing cycle or something else entirely. I'm not an engineer; I'm a COBOL Programmer, so I just use the tools I'm given, but I also don't restrict my view of the horizon.

    ReplyDelete
  4. Interesting comment Old Scout, I'm not an engineer either. It seems as though companies are continually adding cores, threads, new features, and more efficiency to improve performance. It used to be more clockspeed but the megahertz myth debunked that. Others have found, that a solid-state disk circa 2010 provides more performance advantages over a hard drive circa 2010 than upgrading from a upper middle class processor from 2007 to one from 2010 for the average users and power users alike. (tomshardware) So performance increases dont just come from the processor, too. Granted SSDs are expensive per GB

    ReplyDelete