Where are AMD and Intel going in the next few months? - Hardware

This is a discussion on Where are AMD and Intel going in the next few months? - Hardware ; Any knowledge of new things coming out in the next 3 or 4 months from AMD or Intel? Just wondering if there is any reason to wait for further developments, or go ahead and buy a woodcrest or conroe now? ...

+ Reply to Thread
Results 1 to 5 of 5

Thread: Where are AMD and Intel going in the next few months?

  1. Where are AMD and Intel going in the next few months?

    Any knowledge of new things coming out in the next 3 or 4 months from
    AMD or Intel? Just wondering if there is any reason to wait for further
    developments, or go ahead and buy a woodcrest or conroe now?

    Many thanks for any advice...


  2. Re: Where are AMD and Intel going in the next few months?

    On Sat, 02 Dec 2006 13:46:40 -0800, kroger@princeton.edu wrote:

    > Any knowledge of new things coming out in the next 3 or 4 months from
    > AMD or Intel? Just wondering if there is any reason to wait for further
    > developments, or go ahead and buy a woodcrest or conroe now?
    >
    > Many thanks for any advice...


    The next big introductions are next summer when the true quad cores will
    be introduced. Intel already has a "quad" core but it's just a pair of
    dual core chips in the same package which share the same FSB. Next
    summer's chips will be designed to be quad cores so they'll have adequate
    memory bandwidth, if you want to four cores today then get a Woodcrest or
    an Opteron system and skip the QX6700.

  3. Re: Where are AMD and Intel going in the next few months?


    General Schvantzkoph wrote:
    > On Sat, 02 Dec 2006 13:46:40 -0800, kroger@princeton.edu wrote:
    >
    > > Any knowledge of new things coming out in the next 3 or 4 months from
    > > AMD or Intel? Just wondering if there is any reason to wait for further
    > > developments, or go ahead and buy a woodcrest or conroe now?
    > >
    > > Many thanks for any advice...

    >
    > The next big introductions are next summer when the true quad cores will
    > be introduced. Intel already has a "quad" core but it's just a pair of
    > dual core chips in the same package which share the same FSB. Next
    > summer's chips will be designed to be quad cores so they'll have adequate
    > memory bandwidth, if you want to four cores today then get a Woodcrest or
    > an Opteron system and skip the QX6700.



    I was just about to comment on the bandwidth problems with Intel chips
    and wondering why folks are excited about more cores when they are
    hobbled at using them. Well, according to what I've read anyway, I'm
    not an expert.

    It's interesting to me that on the Tom's Hardware site the Woodcrests
    considerably outperformed Opterons, except in one set of tests: memory
    bandwidth. If one is using very large amounts of RAM and heavily
    accessing it, it seems like no matter how much faster the Intels can
    compute something residing in small memory, if the computations require
    rapid exchange with memory such as processing very large quantities of
    time series data, the Opteron's memory bandwidth will trump it. But the
    article barely mentioned that the Opteron beat the Woodcrests in this
    regard. I suspect that none of the benchmarks really simulate what a
    scientist might really do. They don't just run empty computations, they
    usually have to apply them to huge masses of data, from astronomy, to
    biology, and so on. That the Opteron can still beat the Woodcrest when
    accessing RAM makes me hesitant to be confident that getting an Intel
    is neessarily the right move at the moment.

    So, what do you know about the improved bandwidth on future Intel's,
    such as the quads?


  4. Re: Where are AMD and Intel going in the next few months?

    On 2006-12-06, kroger@princeton.edu wrote:
    >
    > It's interesting to me that on the Tom's Hardware site the Woodcrests
    > considerably outperformed Opterons, except in one set of tests: memory
    > bandwidth. If one is using very large amounts of RAM and heavily
    > accessing it, it seems like no matter how much faster the Intels can
    > compute something residing in small memory, if the computations require
    > rapid exchange with memory such as processing very large quantities of
    > time series data, the Opteron's memory bandwidth will trump it. But the
    > article barely mentioned that the Opteron beat the Woodcrests in this
    > regard. I suspect that none of the benchmarks really simulate what a
    > scientist might really do. They don't just run empty computations, they
    > usually have to apply them to huge masses of data, from astronomy, to
    > biology, and so on. That the Opteron can still beat the Woodcrest when
    > accessing RAM makes me hesitant to be confident that getting an Intel
    > is neessarily the right move at the moment.


    Here are some real world numbers for dual core chips doing the types of work
    the labs I admin for actually do:

    http://www.duke.edu/~jlb17/optxeon.pdf

    DYNA is a FEM program, and the Woodcrest Xeons kick Opteron's butt at it.
    This is true even for implicitly solved thermal simulations, which are
    essentially one gigantic matrix inversion done in memory. The conventional
    wisdom on this is that the Woodcrest's gigantic cache is drowning out the
    effects of its subpar memory architecture. Only in a *very* memory
    intensive Matlab simulation run on all 4 cores does the Opteron come out on
    top. Given those results, I just bought a round dual-core Xeons for our
    latest round of cluster upgrades. Quad-core Xeons weren't interesting to me
    because the clockspeeds are so low.

    All the usual cavets apply, mainly YMMV.

    --
    Joshua Baker-LePain
    Department of Biomedical Engineering
    Duke University

  5. Re: Where are AMD and Intel going in the next few months?


    Joshua Baker-LePain wrote:
    > On 2006-12-06, kroger@princeton.edu wrote:
    > >
    > > It's interesting to me that on the Tom's Hardware site the Woodcrests
    > > considerably outperformed Opterons, except in one set of tests: memory
    > > bandwidth. If one is using very large amounts of RAM and heavily
    > > accessing it, it seems like no matter how much faster the Intels can
    > > compute something residing in small memory, if the computations require
    > > rapid exchange with memory such as processing very large quantities of
    > > time series data, the Opteron's memory bandwidth will trump it. But the
    > > article barely mentioned that the Opteron beat the Woodcrests in this
    > > regard. I suspect that none of the benchmarks really simulate what a
    > > scientist might really do. They don't just run empty computations, they
    > > usually have to apply them to huge masses of data, from astronomy, to
    > > biology, and so on. That the Opteron can still beat the Woodcrest when
    > > accessing RAM makes me hesitant to be confident that getting an Intel
    > > is neessarily the right move at the moment.

    >
    > Here are some real world numbers for dual core chips doing the types of work
    > the labs I admin for actually do:
    >
    > http://www.duke.edu/~jlb17/optxeon.pdf
    >
    > DYNA is a FEM program, and the Woodcrest Xeons kick Opteron's butt at it.
    > This is true even for implicitly solved thermal simulations, which are
    > essentially one gigantic matrix inversion done in memory. The conventional
    > wisdom on this is that the Woodcrest's gigantic cache is drowning out the
    > effects of its subpar memory architecture. Only in a *very* memory
    > intensive Matlab simulation run on all 4 cores does the Opteron come out on
    > top. Given those results, I just bought a round dual-core Xeons for our
    > latest round of cluster upgrades. Quad-core Xeons weren't interesting to me
    > because the clockspeeds are so low.
    >
    > All the usual cavets apply, mainly YMMV.
    >
    > --
    > Joshua Baker-LePain
    > Department of Biomedical Engineering
    > Duke University



    Thanks very much. That's just the sort of thing I need. I'm holding off
    on a purchase to see if any more desirable chips come out but after
    some discussions here I think they won't.

    Many thanks...


+ Reply to Thread