Re: 10Ghz CPUs

Discussion in 'Computer Information' started by Paul, Nov 30, 2008.

  1. Paul

    Paul Guest

    John wrote:
    > Why don't we have 10Ghz CPUs as standard in computers today?
    >
    > We seem to have been stuck around the 3Ghz mark give or take 13% or so
    > for some time now. Why haven't processor speeds gone upwards for
    > consumers? And why do we continue to see more and more multi-cores
    > coming to the market?
    >
    > Surely at some point you have to say enough is enough and there is
    > only so many cores any individual is going to need for a system? Some
    > people may only need one core and just require much better speeds than
    > 3.4Ghz. We seem to be stuck in a speed bottle neck.
    >
    > Anyone who knows how long it's going to be before we see a jump in
    > processor speeds I'd love to know.
    >
    > Thanks,
    >
    > John


    Years ago, my company had in-house technology that ran at 10GHz.
    I was sent on a CAD course, to design using that technology, but
    no need ever arose to use it (at least for me).

    What limitations did it have ? The chip could contain about 2000 gates.
    Modern processors have hundreds of millions of transistors in them.
    So the difference in capacities is quite striking. Even a
    relatively simple design, like a PowerPC core (sometimes
    embedded "for free" in some FPGA technologies), would be
    too much for a single chip of the 10GHz technology.

    One of our competitors had a version of that technology,
    that might have run some kinds of circuits at 40GHz. But
    again, it would have similar limitations. You can only
    do the simplest of data transfer functions with stuff like
    that. And it also ran hot. You'd be surprised how high
    a voltage it needed to operate.

    The transistors inside those devices, are relatively large
    by comparison to Intel's current 45nm designs. I specifically
    asked, while taking my CAD course, whether the company would be
    able to scale the transistors down, allowing larger density
    chips to be built. The answer I got, is they were as small
    as they were ever going to get.

    Now, assume they were wrong, and with the usual Moore's law
    improvements, a smaller transistor could be made. Say there
    were enough transistors for a simple RISC processor.

    If the core runs at 10GHz, it has to be fed by a memory of
    some kind. The ratio between core speed, and memory subsystem
    speed, is already ridiculous. It is made tenable, by more
    than one layer of caching. The size of each level of caching
    has some threshold, below which things really start to slow
    down. There is no point running at 10GHz, if you have to
    wait 1000 cycles for something to come from some level of
    cache. Like, if I wanted 32KB or 64KB of cache for an
    L1, again, it might take multiple chips just to give me
    that much cache. Then I'd need microwave wiring to connect
    the chips together. (Driving a 10GHz signal through
    FR-4 is not easy.)

    The beauty of CMOS, is it has continued to scale for so many
    years. At one time, the features were large enough, you
    could view them with a child's microscope or even a
    magnifying glass. It's a miracle that it made it
    to 3GHz. I can remember a time, when if you said
    "3 GHz", the instant thought process would be
    "that'll take GaAs". But now, CMOS routinely fills
    that role.

    There is still a usage for the ultra-high speed technologies.
    For example, to operate high speed fiber optic trunks between
    cities, the signal coming from the receiver, has to be
    reduced in speed and made more parallel, to be handled
    by other layers of circuitry. For those applications,
    some of that 100GHz Ft technology is still the right answer.

    HTH,
    Paul
     
    Paul, Nov 30, 2008
    #1
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. Jimmy Dean
    Replies:
    0
    Views:
    461
    Jimmy Dean
    Oct 30, 2003
  2. Scratch

    What's CPUs' (P4 I think) HT function?

    Scratch, Jan 6, 2004, in forum: Computer Support
    Replies:
    1
    Views:
    432
  3. mhicaoidh

    Dual CPUs split load

    mhicaoidh, Jul 15, 2005, in forum: Computer Support
    Replies:
    2
    Views:
    435
    Brian
    Jul 15, 2005
  4. pickled

    Re: 10Ghz CPUs

    pickled, Nov 30, 2008, in forum: Computer Information
    Replies:
    1
    Views:
    476
    - Bobb -
    Dec 10, 2008
  5. computergeek3000

    Re: 10Ghz CPUs

    computergeek3000, Dec 8, 2008, in forum: Computer Information
    Replies:
    0
    Views:
    429
    computergeek3000
    Dec 8, 2008
Loading...

Share This Page