Intel's x86-64

Discussion in 'NZ Computing' started by Adam Warner, Feb 18, 2004.

  1. Adam Warner

    Adam Warner Guest

    Hi all,

    The big news to come out of the Intel Developer Forum is that Intel will
    be effectively producing AMD64 compatible processors. And soon.
    <http://theinquirer.net/?article=14192> "In a Q&A, [Craig Barrett] said
    that while Intel and AMD's architectures were quite different, and the
    real question is whether OSes and other software ran on the two families.
    "Broadly, he said, for the most part the OSes for one will run on
    another."

    Intel has ceded the x86 Instruction Set Architecture (ISA) to AMD. In the
    future you almost certainly won't be running IA64. That is the Itanium
    architecture.

    The 64-bit instruction set architecture you are likely to soon be using is
    called AMD64. However when talking about AMD and Intel processors that
    implement the AMD64 ISA the generic term x86-64 will probably be helpful.

    Intel is continuing to downplay the value of their upcoming x86-64
    processors merely calling them a "memory extension technology" and IA-32e.
    This is an attempt to salvage a position for their Itanium architecture
    and imply that AMD64 is an inferior design.

    AMD64 is a well designed 64-bit ISA that is the simply the product of good
    and backwards compatible engineering. Consider how inferior an x86-64 ISA
    could have been if Intel had controlled its design while attempting to
    position it as inferior to the Itanium. AMD had the self interest to
    create the best possible x86-64 processor. Intel still lacks part of this
    self interest and will do so until they abandon the Itanium as their
    high-end platform.

    Of course Intel will include incompatible extensions to x86-64:
    <http://theregister.co.uk/content/61/35628.html> "Intel has some (things)
    unique to Intel, which we will make sure people write, port and tune to."

    We're all used to this kind of scenario with i386 being the base
    compatible architecture with extensions being built upon it in a one-up
    fashion.

    Welcome to the world of x86-64. Designed by AMD.

    Regards,
    Adam
    Adam Warner, Feb 18, 2004
    #1
    1. Advertising

  2. Adam Warner

    Adam Warner Guest

    Reference specifications are linked from here:
    <http://developer.intel.com/technology/64bitextensions/>

    The 64-bit extension technology is an Intel innovation. While AMD is not
    mentioned in the summary the Intel "Processors with 64-bit extension
    technology will support 64-bit extended operating systems from Microsoft,
    Red Hat and SuSE." Which means they must be AMD64 compatible.

    Intel is claiming their Xeon range will gain 64-bit support this year.
    This may be an optimistic timeline in order to dissuade customers from
    migrating to the AMD Opteron.

    Having just searched through the 684 pages of the IA-32e specifications
    Intel appears to has achieved at least one extraordinary feat: not
    mentioning AMD once. My cursory glance is that everything is exactly the
    same as AMD64 with the addition of Intel's SSE3 instructions which came
    out with Prescott (AMD's current 64-bit processors support up to SSE2).

    Regards,
    Adam
    Adam Warner, Feb 18, 2004
    #2
    1. Advertising

  3. "Adam Warner" <> wrote in message
    news:p...
    > Hi all,
    >
    > The big news to come out of the Intel Developer Forum is that Intel will
    > be effectively producing AMD64 compatible processors. And soon.
    > <http://theinquirer.net/?article=14192> "In a Q&A, [Craig Barrett] said
    > that while Intel and AMD's architectures were quite different, and the
    > real question is whether OSes and other software ran on the two families.
    > "Broadly, he said, for the most part the OSes for one will run on
    > another."
    >
    > Intel has ceded the x86 Instruction Set Architecture (ISA) to AMD. In the



    From what I remember AMD lost the court case with Intel a few years back,
    this effectively gave Intel the rights to x86-64 at no charge
    Nathan Mercer, Feb 18, 2004
    #3
  4. Adam Warner

    David Preece Guest

    Nathan Mercer wrote:
    > From what I remember AMD lost the court case with Intel a few years back,
    > this effectively gave Intel the rights to x86-64 at no charge


    I read something about that as well. Not too sure what it was all about
    though.

    But we have to ask, does it matter? Has AMD successfully shaken off it's
    "no good for servers" image? Do they have a sustainable technology lead?
    Will Intel and Microsoft conspire to ensure 64 bit windows goes like a
    dog on AMD?

    My money is yes for fixing their image; yes for a sustainable technology
    lead (their processors are a hell of a lot faster on any metric, per
    clock, per watt, per watt of heat, per dollar or per square millimetre);
    probably no for Intel and Microsoft openly conspiring, although I
    think we can assume that Microsoft's C++ compiler will remain tuned to
    the P4's long pipelines; and that none of this will matter and AMD will
    remain number two.

    More interested to see what happens with Sun and Opteron to be honest.

    Personally I'm monster impressed by the new 970FX's which are faster on
    all metrics *again*. Apparently 24W at 2GHz - or roughly the same amount
    of computing "grunt" as a 3.4GHz P4 for a third as much heat.

    Dave
    David Preece, Feb 18, 2004
    #4
  5. Suddenly, David Preece sprang forth and uttered these pithy words:
    > Personally I'm monster impressed by the new 970FX's which are faster on
    > all metrics *again*. Apparently 24W at 2GHz - or roughly the same amount
    > of computing "grunt" as a 3.4GHz P4 for a third as much heat.


    Yes, seems like at long last the theoritical benefits of [pure-ish] RISC
    are really coming in...
    --
    aaronl at consultant dot com
    For every expert, there is an equal and
    opposite expert. - Arthur C. Clarke
    Aaron Lawrence, Feb 18, 2004
    #5
  6. Adam Warner wrote:
    > The 64-bit extension technology is an Intel innovation. While AMD is not
    > mentioned in the summary the Intel "Processors with 64-bit extension
    > technology will support 64-bit extended operating systems from Microsoft,
    > Red Hat and SuSE." Which means they must be AMD64 compatible.


    Not necessarily true... these three OS companies also have Itanium
    versions of their software don't they? Maybe they're shoving their
    Itanium shit into these CPU's?

    --
    Dave Hall
    http://Dave.net.nz
    We have Hangman, Pacman, and Space Invaders
    T.N.O. - Dave.net.nz, Feb 18, 2004
    #6
  7. David Preece wrote:
    > But we have to ask, does it matter?


    Don't think so.

    > Has AMD successfully shaken off it's
    > "no good for servers" image?


    I know plenty of places that are evaluating them... don't know any that
    have chosen them yet, but hey, they weren't even evaluating them prior
    to Opterons.

    > Do they have a sustainable technology lead?


    I dont think it can be called a "sustainable technology lead" but it
    certainly is an interesting technology.

    > Will Intel and Microsoft conspire to ensure 64 bit windows goes like a
    > dog on AMD?


    I dont think so... it doesnt help MS at all if their OS runs like a dog.

    > My money is yes for fixing their image; yes for a sustainable technology
    > lead (their processors are a hell of a lot faster on any metric, per
    > clock, per watt, per watt of heat, per dollar or per square millimetre);
    > probably no for Intel and Microsoft openly conspiring, although I think
    > we can assume that Microsoft's C++ compiler will remain tuned to the
    > P4's long pipelines; and that none of this will matter and AMD will
    > remain number two.


    Mostly agreee with you and your reasoning.

    > More interested to see what happens with Sun and Opteron to be honest.


    Very interesting yes.

    > Personally I'm monster impressed by the new 970FX's which are faster on
    > all metrics *again*. Apparently 24W at 2GHz - or roughly the same amount
    > of computing "grunt" as a 3.4GHz P4 for a third as much heat.


    Havent read that... might go to google.

    --
    Dave Hall
    http://Dave.net.nz
    We have Hangman, Pacman, and Space Invaders
    T.N.O. - Dave.net.nz, Feb 18, 2004
    #7
  8. Adam Warner

    Adam Warner Guest

    Hi Dave,

    > Adam Warner wrote:
    >> The 64-bit extension technology is an Intel innovation. While AMD is
    >> not mentioned in the summary the Intel "Processors with 64-bit
    >> extension technology will support 64-bit extended operating systems
    >> from Microsoft, Red Hat and SuSE." Which means they must be AMD64
    >> compatible.

    >
    > Not necessarily true... these three OS companies also have Itanium
    > versions of their software don't they? Maybe they're shoving their
    > Itanium shit into these CPU's?


    My final paragraph made it plain that this is the AMD64 ISA. I 100%
    approve of Intel copying AMD's ISA. I simply believe that an appropriate
    level of attribution hasn't occurred [we would demand this in individual
    relationships. Is it OK for companies "to appropriate without due
    acknowledgement (the ideas or expressions of another)."? (1913 Webster)]

    This is a link to a compatibility discussion on the Linux kernel mailing
    list: <http://marc.theaimsgroup.com/?l=linux-kernel&m=107706941309340&w=2>

    Regards,
    Adam
    Adam Warner, Feb 18, 2004
    #8
  9. Adam Warner wrote:
    >>Not necessarily true... these three OS companies also have Itanium
    >>versions of their software don't they? Maybe they're shoving their
    >>Itanium shit into these CPU's?


    > My final paragraph made it plain that this is the AMD64 ISA. I 100%
    > approve of Intel copying AMD's ISA. I simply believe that an appropriate
    > level of attribution hasn't occurred [we would demand this in individual
    > relationships. Is it OK for companies "to appropriate without due
    > acknowledgement (the ideas or expressions of another)."? (1913 Webster)]
    > This is a link to a compatibility discussion on the Linux kernel mailing
    > list: <http://marc.theaimsgroup.com/?l=linux-kernel&m=107706941309340&w=2>


    Didn't AMD make their 64bit extensions open so that others could do it?

    I'm sure I read about transmeta making the first virtual AMD64...before
    AMD had final silicon.

    Found a similar article, from around the same time frame.

    http://www.theinquirer.net/?article=2396


    --
    Dave Hall
    http://Dave.net.nz
    We have Hangman, Pacman, and Space Invaders
    T.N.O. - Dave.net.nz, Feb 18, 2004
    #9
  10. Adam Warner

    Adam Warner Guest

    Hi Dave,

    >> My final paragraph made it plain that this is the AMD64 ISA. I 100%
    >> approve of Intel copying AMD's ISA. I simply believe that an
    >> appropriate level of attribution hasn't occurred [we would demand this
    >> in individual relationships. Is it OK for companies "to appropriate
    >> without due acknowledgement (the ideas or expressions of another)."?
    >> (1913 Webster)] This is a link to a compatibility discussion on the
    >> Linux kernel mailing list:
    >> <http://marc.theaimsgroup.com/?l=linux-kernel&m=107706941309340&w=2>

    >
    > Didn't AMD make their 64bit extensions open so that others could do it?
    >
    > I'm sure I read about transmeta making the first virtual AMD64...before
    > AMD had final silicon.
    >
    > Found a similar article, from around the same time frame.
    >
    > http://www.theinquirer.net/?article=2396


    OK. Now link to Intel's acknowledgement that they are implementing the
    AMD64 Instruction Set Architecture.

    Regards,
    Adam
    Adam Warner, Feb 18, 2004
    #10
  11. Adam Warner wrote:
    >>I'm sure I read about transmeta making the first virtual AMD64...before
    >>AMD had final silicon.
    >>Found a similar article, from around the same time frame.
    >>http://www.theinquirer.net/?article=2396


    > OK. Now link to Intel's acknowledgement that they are implementing the
    > AMD64 Instruction Set Architecture.


    The best I can do while at work is this.
    http://www.theinquirer.net/?article=14040
    a heap of links from ages ago to just recently with quotes from many
    intel execs mentioning many things that "don't quite confirm" it, but
    get very close.

    I doubt that we will ever get anyone from intel to admit that.

    --
    Dave Hall
    http://Dave.net.nz
    We have Hangman, Pacman, and Space Invaders
    T.N.O. - Dave.net.nz, Feb 19, 2004
    #11
  12. Adam Warner

    Warwick Guest

    On Wed, 18 Feb 2004 15:38:16 +1300, Adam Warner wrote:

    > Hi all,
    >
    > The big news to come out of the Intel Developer Forum is that Intel will
    > be effectively producing AMD64 compatible processors. And soon.
    > <http://theinquirer.net/?article=14192> "In a Q&A, [Craig Barrett] said
    > that while Intel and AMD's architectures were quite different, and the
    > real question is whether OSes and other software ran on the two families.
    > "Broadly, he said, for the most part the OSes for one will run on
    > another."
    >
    > Intel has ceded the x86 Instruction Set Architecture (ISA) to AMD. In the
    > future you almost certainly won't be running IA64. That is the Itanium
    > architecture.
    >
    > The 64-bit instruction set architecture you are likely to soon be using is
    > called AMD64. However when talking about AMD and Intel processors that
    > implement the AMD64 ISA the generic term x86-64 will probably be helpful.
    >
    > Intel is continuing to downplay the value of their upcoming x86-64
    > processors merely calling them a "memory extension technology" and IA-32e.
    > This is an attempt to salvage a position for their Itanium architecture
    > and imply that AMD64 is an inferior design.
    >
    > AMD64 is a well designed 64-bit ISA that is the simply the product of good
    > and backwards compatible engineering. Consider how inferior an x86-64 ISA
    > could have been if Intel had controlled its design while attempting to
    > position it as inferior to the Itanium. AMD had the self interest to
    > create the best possible x86-64 processor. Intel still lacks part of this
    > self interest and will do so until they abandon the Itanium as their
    > high-end platform.
    >
    > Of course Intel will include incompatible extensions to x86-64:
    > <http://theregister.co.uk/content/61/35628.html> "Intel has some (things)
    > unique to Intel, which we will make sure people write, port and tune to."
    >
    > We're all used to this kind of scenario with i386 being the base
    > compatible architecture with extensions being built upon it in a one-up
    > fashion.
    >
    > Welcome to the world of x86-64. Designed by AMD.
    >
    > Regards,
    > Adam


    A question please....
    Can C/C++ code be recompiled for 64 bit processors without changing the
    source?
    If so, how much speed is gained?
    Do any dev packages support 64 bit compilers?
    Have they changed the size of double and long ?

    This might be the wrong place to ask such questions, if that is the case
    sorry:)

    cheers
    Warwick
    Warwick, Feb 19, 2004
    #12
  13. Adam Warner

    Adam Warner Guest

    Hi Dave,

    >> OK. Now link to Intel's acknowledgement that they are implementing the
    >> AMD64 Instruction Set Architecture.


    [...]

    > I doubt that we will ever get anyone from intel to admit that.


    Exactly. We have to infer that Intel is implementing the AMD64 ISA by
    comparing Intel's IA-32e ISA to the AMD64 ISA and discovering that they
    are effectively identical. Intel could have acknowledged this by supplying
    a short document outlining the differences but then the facade that this
    is Intel innovation couldn't be sustained.

    BTW this appears to be the best description of the differences available
    to date, courtesy of Zalman Stern in comp.arch:
    <http://groups.google.co.nz/groups?selm=2d24c5a8.0402181022.e03d706%40posting.google.com>

    "Intel has clearly done what AMD has been doing all along: architected for
    compatibility with another company's architecture."

    The amazing news of the last two days is that Intel is transitioning to a
    company that produces AMD-compatible processors. I have little doubt Intel
    will market its clones so well that most of the world will soon believe
    that Intel is and always has been the x86-64 innovator. I have little
    doubt most people will forget the outrageous statements of Intel
    executives, as discussed just under one year ago in nz.comp:
    <http://groups.google.co.nz/groups?selm=pan.2003.02.25.00.56.56.699413%40consulting.net.nz>

    Regards,
    Adam
    Adam Warner, Feb 19, 2004
    #13
  14. "Adam Warner" <> wrote in message
    news:p...
    > Hi all,
    >
    > The big news to come out of the Intel Developer Forum is that Intel will
    > be effectively producing AMD64 compatible processors. And soon.
    > <http://theinquirer.net/?article=14192> "In a Q&A, [Craig Barrett] said
    > that while Intel and AMD's architectures were quite different, and the
    > real question is whether OSes and other software ran on the two families.
    > "Broadly, he said, for the most part the OSes for one will run on
    > another."
    >
    > Intel has ceded the x86 Instruction Set Architecture (ISA) to AMD. In the
    > future you almost certainly won't be running IA64. That is the Itanium
    > architecture.


    We are all so lucky. Lucky lucky lucky. Twenty years of x86 kludges get a 64-bit face-lift, and now the
    architecture looks to be with us unto death. Whoo-bloody-hoo. Thanks AMD. Thanks Intel. You immense piles
    of steaming smeg.

    Tony.
    The Black Wibble, Feb 19, 2004
    #14
  15. The Black Wibble wrote:
    >>Intel has ceded the x86 Instruction Set Architecture (ISA) to AMD. In the
    >>future you almost certainly won't be running IA64. That is the Itanium
    >>architecture.


    > We are all so lucky. Lucky lucky lucky. Twenty years of x86 kludges get a 64-bit face-lift, and now the
    > architecture looks to be with us unto death. Whoo-bloody-hoo. Thanks AMD. Thanks Intel. You immense piles
    > of steaming smeg.


    Whats so bad about it all?

    --
    Dave Hall
    http://Dave.net.nz
    We have Hangman, Pacman, and Space Invaders
    T.N.O. - Dave.net.nz, Feb 19, 2004
    #15
  16. Adam Warner

    Adam Warner Guest

    Hi Warwick,

    > A question please....
    > Can C/C++ code be recompiled for 64 bit processors without changing the
    > source?


    If it can the source is termed "64-bit clean".

    > If so, how much speed is gained?


    All other things being equal a transition to 64-bit applications typically
    leads to slowdowns. But the x86-64 architecture is special in that it also
    improves upon the x86-32 architecture in other ways. The main improvement
    is a doubling of the number of general purpose registers. x86-64
    implements sixteen 64-bit general purpose registers.

    A compiler supporting the architecture will be able to assemble extra
    intermediate calculations without having to store them in main memory,
    leading to overall faster execution.

    Most applications tend to run faster. Some applications involving very big
    numbers can even run four times faster: <http://www.swox.com/gmp/32vs64.html>

    Note of course that we are yet to see any Intel's x86-64 benchmarks.

    Compilation for AMD64 is already well supported--with the notable
    exception of Intel, but that will surely change (perhaps by also modifying
    Intel's binaries to disable a CPUID check).

    > Do any dev packages support 64 bit compilers? Have they changed the size
    > of double and long ?


    For GCC:
    <http://gcc.gnu.org/onlinedocs/gcc-3.3.1/gcc/i386-and-x86-64-Options.html>

    -m32
    -m64
    Generate code for a 32-bit or 64-bit environment. The 32-bit
    environment sets int, long and pointer to 32 bits and generates
    code that runs on any i386 system. The 64-bit environment sets int
    to 32 bits and long and pointer to 64 bits and generates code for
    AMD's x86-64 architecture.

    Regards,
    Adam
    Adam Warner, Feb 19, 2004
    #16
  17. Adam Warner wrote:
    > Most applications tend to run faster. Some applications involving very big
    > numbers can even run four times faster: <http://www.swox.com/gmp/32vs64.html>
    > Note of course that we are yet to see any Intel's x86-64 benchmarks.
    > Compilation for AMD64 is already well supported--with the notable
    > exception of Intel, but that will surely change (perhaps by also modifying
    > Intel's binaries to disable a CPUID check).


    This has already happened, read it the other day.

    --
    Dave Hall
    http://Dave.net.nz
    We have Hangman, Pacman, and Space Invaders
    T.N.O. - Dave.net.nz, Feb 19, 2004
    #17
  18. "T.N.O. - Dave.net.nz" <> wrote in message
    news:c1147s$1ck1cc$-berlin.de...
    > The Black Wibble wrote:
    > >>Intel has ceded the x86 Instruction Set Architecture (ISA) to AMD. In the
    > >>future you almost certainly won't be running IA64. That is the Itanium
    > >>architecture.

    >
    > > We are all so lucky. Lucky lucky lucky. Twenty years of x86 kludges get a 64-bit face-lift, and now the
    > > architecture looks to be with us unto death. Whoo-bloody-hoo. Thanks AMD. Thanks Intel. You immense

    piles
    > > of steaming smeg.

    >
    > Whats so bad about it all?


    Oh, the gruesome instruction set and the memory bandwidth hungry code produced from it. But that small
    outburst was the remnants of my MC68000 programming past surfacing. I feel better now having let it out. RIP
    MC68000.

    Tony.
    The Black Wibble, Feb 19, 2004
    #18
  19. Adam Warner

    AD. Guest

    On Thu, 19 Feb 2004 14:32:28 +1300, The Black Wibble wrote:

    > We are all so lucky. Lucky lucky lucky. Twenty years of x86 kludges get
    > a 64-bit face-lift, and now the architecture looks to be with us unto
    > death. Whoo-bloody-hoo. Thanks AMD. Thanks Intel. You immense piles of
    > steaming smeg.


    That;s overstating it, AMD has removed a bunch of legacy cruft from the
    64bit mode(s). A lot of the original criticism of the x86 arch doesn't
    really apply anymore (not all of course though).

    Anyway, IMO it only really affects assembly programmers etc.

    Slight change of topic...

    Maybe Intel brought out IA64 (Itanium) to make programmers appreciate IA32? ;)

    And someone joked on another newsgroup that the 64bit Xeon should be
    called the 'Iberg' because it sinks the 'Itanic' hehe

    Cheers
    Anton
    AD., Feb 19, 2004
    #19
  20. "AD." <> wrote in message news:p...
    > On Thu, 19 Feb 2004 14:32:28 +1300, The Black Wibble wrote:
    >
    > > We are all so lucky. Lucky lucky lucky. Twenty years of x86 kludges get
    > > a 64-bit face-lift, and now the architecture looks to be with us unto
    > > death. Whoo-bloody-hoo. Thanks AMD. Thanks Intel. You immense piles of
    > > steaming smeg.

    >
    > That;s overstating it, AMD has removed a bunch of legacy cruft from the
    > 64bit mode(s). A lot of the original criticism of the x86 arch doesn't
    > really apply anymore (not all of course though).
    >
    > Anyway, IMO it only really affects assembly programmers etc.
    >
    > Slight change of topic...
    >
    > Maybe Intel brought out IA64 (Itanium) to make programmers appreciate IA32? ;)


    The IA-64 has 128 general purpose registers, 128 floating point registers, and the stack frames can be
    allocated on blocks of general purpose registers instead of cache or main memory except as a last resort.
    Arithmetic instructions with the format: opcode r(n)=r(n), r(n). I could go for that. The nasty thing about
    the Itanium is that the assembly programmer has to write code that obeys the parallel nature of the execution
    engines making the CPU hard to program, the instructions looking out of order, and the source code hard to
    read. How sad. Never mind. I did not know AMD got rid of some of the old slop. Good job, because it was
    just a waste of transistors. I'd like to see the entirety of the ancient segment:eek:ffset support dropped, and
    the mnemonics prettied up, made orthogonal and straightfoward. None of it is important for 99% of
    programmers. I just like things kept simple and well designed.

    > And someone joked on another newsgroup that the 64bit Xeon should be
    > called the 'Iberg' because it sinks the 'Itanic' hehe


    RIP Itanium. I hope not too many people get caught up in its wake.

    Tony.
    The Black Wibble, Feb 19, 2004
    #20
    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. roberthob

    an Intel PRO/DSL 2100 or Intel PRO/DSL 2200 Modem

    roberthob, Sep 25, 2005, in forum: Computer Support
    Replies:
    0
    Views:
    1,541
    roberthob
    Sep 25, 2005
  2. Uplink

    x86 and intel

    Uplink, Feb 2, 2006, in forum: Computer Information
    Replies:
    3
    Views:
    355
    Uplink
    Feb 4, 2006
  3. =?Utf-8?B?RWxsaW90IEh1ZGdpbnM=?=

    Why is there an x86 emu if a processor is x86-64?

    =?Utf-8?B?RWxsaW90IEh1ZGdpbnM=?=, Jul 23, 2006, in forum: Windows 64bit
    Replies:
    4
    Views:
    627
  4. Daniel
    Replies:
    11
    Views:
    588
  5. markm75
    Replies:
    7
    Views:
    1,067
    S.SubZero
    Jan 9, 2008
Loading...

Share This Page