x86 suddenly has a life ahead of it? WTF!?

Posted:
in Future Apple Hardware edited January 2014
I'm a little confused or torn by the switch to Intel. I'm hoping it will be good in the long run, but I still have these questions (which I'm sure have been asked a million time already)...



1) Is the x86 arch really antient/legacy?

2) Is the x86 arch really hitting a laws-of-physics brick wall?

3) Why the hell would Apple move us to an ancient arch?

4) Does Intel have multi-core-multi-proc in the roadmap?

5) I haven't heard anything about 64bit Intel for Mac? WTF?!

6) Will the CELL proc take off and leave Apple in the dust - with MS being the only ones with the resources to switch procs without killing themselves?

7) Is AMD x86, such that if Apple in the future got shafted by intel they'd be able to use AMD without having to take another 5 years switch/port over?

8) On one hand I like the idea of being able to run Windows apps within OS X, but how is Apple going to keep devs making native OS X apps?

9) How many years does the x86 arch have left? Will we have to make another switch in 5 years when some genius comes up with a GAZILLION-SUPER-DUPER-TERAFLOP-MEGA-SUPER-CHIP?

10) Where can I find an official Intel roadmap? Anyone have a link to one???
«13

Comments

  • Reply 1 of 49
    ruudruud Posts: 20member
    Quote:

    Originally posted by Macvault

    I'm a little confused or torn by the switch to Intel. I'm hoping it will be good in the long run, but I still have these questions (which I'm sure have been asked a million time already)...



    One by one:



    1) Is the x86 arch really antient/legacy?



    The instruction set dates back to the 8086/8088 processors used in the very first IBM PC's in the early 80's. That doesn't mean that the processor design itself is ancient.



    2) Is the x86 arch really hitting a laws-of-physics brick wall?



    No. Laws of physics apply equally to all processor architectures.



    3) Why the hell would Apple move us to an ancient arch?



    Apple isn't.



    4) Does Intel have multi-core-multi-proc in the roadmap?



    Yes.



    5) I haven't heard anything about 64bit Intel for Mac? WTF?!



    Probably because the powerbooks are the first machines scheduled to get the x86 processors. Current P4's and planned Pentium M's do have 64-bit capability, though.



    6) Will the CELL proc take off and leave Apple in the dust - with MS being the only ones with the resources to switch procs without killing themselves?



    The Cell processor is not intended as a general computing CPU, it's only efficient for multimedia/gaming type applciations.



    7) Is AMD x86, such that if Apple in the future got shafted by intel they'd be able to use AMD without having to take another 5 years switch/port over?



    Yes.



    8 ) On one hand I like the idea of being able to run Windows apps within OS X, but how is Apple going to keep devs making native OS X apps?



    You won't be able to run Windows apps within OS X out of the box, you'd still need something like Virtual PC, VMware or WINE.



    9) How many years does the x86 arch have left? Will we have to make another switch in 5 years when some genius comes up with a GAZILLION-SUPER-DUPER-TERAFLOP-MEGA-SUPER-CHIP?



    Considering the vast majority of the world's computers run on the x86 architecture, that's where most of the research/engineering effort is spent. Even if someone else came up with a radically better design, it's likely that by sheer inertia x86 would eventually come out ahead anyway (Itanium anyone).



    10) Where can I find an official Intel roadmap? Anyone have a link to one???



    The official one: http://www.intel.com/products/roadmap/



    See hardware sites such as Anandtech, Tom's hardware, endian.net etc. for unofficial ones.
  • Reply 2 of 49
    skatmanskatman Posts: 609member
    Macvault,

    you don't get out much, do you?
  • Reply 3 of 49
    powerdocpowerdoc Posts: 8,123member
    According to Pentium roadmap, the netburst architecture will stop. Most new chips, will be pentium M derivated chips and will be multicore. In ten years it's possible to see chip with 32 cores in it.



    The X86 chips are no more CISC. Basically this chips are build around a RISC chip feed by a transcoder who translate X86 instructions into simpler instructions codes.

    The PPC 970 is built the same way, the PPC instructions are translated in a simpler code assimalated by the heart of the chip. The difference there, is that this process of translation from the original PPC code to the power 4 code is a much simple process.

    The translation process of the X86 code is more complicated, and bring a performance penalty.



    If you want more info about this subject, Ars Technica will provide you good articles about how work a chip.



    The brick wall is more a wall in term of speed mhz, than a wall in term of performance. The future of chips is multicore, and that't the way are tooking both AMD and Intel.



    For cell processors, we can just only make guess, but it appears that this chip who are supposed to shine for the game consoles, are not that great for desktop or laptop according to Apple.



    One other key of the switch to intel is the laptops computers. Intel has the best laptop chip : the pentium M. If you consider now, that in the US the laptop market is bigger than the desktop one, you will understand why a good laptop chip is mandatory for Apple.

    Unfortunately IBM totally fucked with the G5 and where not able to make them work in a laptop. Perhaps it's the main reason for the switch. The G5 powerbooks will stay vaporware for ever.
  • Reply 4 of 49
    sillyfoolsillyfool Posts: 35member
    Quote:

    The X86 chips are no more CISC. Basically this chips are build around a RISC chip feed by a transcoder who translate X86 instructions into simpler instructions codes.



    Yes and no. The x86 is not RISC but RISC isn't RISC anymore either. At least 10 years ago the incremental 'imporvements' to RISC starting adding in the kinds of complexity that made many of the distinctions between RISC and CISC more or less academic. Architects acknowledge that we entered the post-RISC era almost a decade ago.



    Having said that, the fact is that the x86 architecture has gone through major reworkings more than once. And undoubtly will go through more. But every step of the way it maintains backwards compatibility with the key x86 features. You can argue that things aren't fully binary compatible all the way back to the stone axe but at some point you just have to let some things fall off the list.
  • Reply 5 of 49
    brussellbrussell Posts: 9,812member
    Question: If there are still some legacy drawbacks to Intel's chips, is it possible that OS X could have an advantage over Windows going into the future with Intel?



    Is it possible that in the future, Windows will have to go through some transition pains that OS X could avoid?
  • Reply 6 of 49
    toweltowel Posts: 1,479member
    Quote:

    Originally posted by sillyfool

    Having said that, the fact is that the x86 architecture has gone through major reworkings more than once. And undoubtly will go through more. But every step of the way it maintains backwards compatibility with the key x86 features. You can argue that things aren't fully binary compatible all the way back to the stone axe but at some point you just have to let some things fall off the list.



    I was thinking about that. Since Apple is Intel's first brand-new high-volume customer since, well, IBM, might they see an opportunity to start lopping off some of that legacy cruft? The Mac platform can be a multi-million unit development showcase for Intel, that could help them overcome the enormous inertia inherent in the Wintel camp. I'm not saying they would develop brand-new chips and technologies just for Apple, but they might see Apple as a wedge to help them deploy new chips and techs they'd like to develop anyway. I've read that a BIOS replacement (EFI?) might be one example. Intel's leadership must hate how their products are always characterized as crufty and inelegant, when the biggest reason is this enormous dead weight of legacy they've been chained to.
  • Reply 7 of 49
    thttht Posts: 3,929member
    Quote:

    Originally posted by BRussell

    Question: If there are still some legacy drawbacks to Intel's chips, is it possible that OS X could have an advantage over Windows going into the future with Intel?



    Is it possible that in the future, Windows will have to go through some transition pains that OS X could avoid?




    Software-wise, Apple and MS are on a level playing field with Intel.



    Apple's advantages are what they have been, Steve Jobs DNA. That is, Apple products will typically have taste and elegance in design, for the most part.



    MS's drawback is that they are so huge that they could suffer from bureaucratic rot and not adjust quickly enough. Like, if Longhorn is sufficiently gargantuan, MS will not be able to support new hardware features as quickly as Apple. Then again, everyone develops for Windows, so it's not a problem.
  • Reply 8 of 49
    powerdocpowerdoc Posts: 8,123member
    Quote:

    Originally posted by sillyfool

    Yes and no. The x86 is not RISC but RISC isn't RISC anymore either. At least 10 years ago the incremental 'imporvements' to RISC starting adding in the kinds of complexity that made many of the distinctions between RISC and CISC more or less academic. Architects acknowledge that we entered the post-RISC era almost a decade ago.



    Having said that, the fact is that the x86 architecture has gone through major reworkings more than once. And undoubtly will go through more. But every step of the way it maintains backwards compatibility with the key x86 features. You can argue that things aren't fully binary compatible all the way back to the stone axe but at some point you just have to let some things fall off the list.




    You are right, I just made a simplification. Today X86 architecture has nothing to do except the instruction code with the original one.
  • Reply 9 of 49
    brussellbrussell Posts: 9,812member
    Quote:

    Originally posted by THT

    Software-wise, Apple and MS are on a level playing field with Intel.



    Apple's advantages are what they have been, Steve Jobs DNA. That is, Apple products will typically have taste and elegance in design, for the most part.



    MS's drawback is that they are so huge that they could suffer from bureaucratic rot and not adjust quickly enough. Like, if Longhorn is sufficiently gargantuan, MS will not be able to support new hardware features as quickly as Apple. Then again, everyone develops for Windows, so it's not a problem.




    The reason I wondered about that is that one often hears about how the Pentium translates the x86 instruction set into a more RISC-like instruction set. Since OS X doesn't have that old x86 legacy, couldn't it bypass that translation?
  • Reply 10 of 49
    powerdocpowerdoc Posts: 8,123member
    Quote:

    Originally posted by BRussell

    The reason I wondered about that is that one often hears about how the Pentium translates the x86 instruction set into a more RISC-like instruction set. Since OS X doesn't have that old x86 legacy, couldn't it bypass that translation?



    I don't think so, because there is no direct communication between the Post-Risc core and the RAM : everything have to be translated.

    Bypassing will recquiere a direct channel.



    In an other way, I don't even know if this architecture is not subject to change at a very high rate. I am nearly sure that the internal code of a pentium M, have nothing to do with the one of a pentium 4.
  • Reply 11 of 49
    vox barbaravox barbara Posts: 2,021member
    Quote:

    Originally posted by Powerdoc

    I don't think so, because there is no direct communication between the Post-Risc core and the RAM : everything have to be translated.

    Bypassing will recquiere a direct channel.



    In an other way, I don't even know if this architecture is not subject to change at a very high rate. I am nearly sure that the internal code of a pentium M, have nothing to do with the one of a pentium 4.




    So it is actually a different class of processors, isn't it?

    Why the name "Pentium"?
  • Reply 12 of 49
    benroethigbenroethig Posts: 2,782member
    Nameplate loyalty.
  • Reply 13 of 49
    thttht Posts: 3,929member
    Quote:

    Originally posted by BRussell

    The reason I wondered about that is that one often hears about how the Pentium translates the x86 instruction set into a more RISC-like instruction set. Since OS X doesn't have that old x86 legacy, couldn't it bypass that translation?



    No.



    The Pentium Pro/II/III/4/M only understand x86 instructions. The micro-ops are irrelevant. The translation has neglegible effects on performance, if not zero. The only thing it costs are transistors. In today's transistors, it's <1% of the total amount of transistors.
  • Reply 14 of 49
    zoranszorans Posts: 187member
  • Reply 15 of 49
    sillyfoolsillyfool Posts: 35member
    ARGH!!!! stupid forum software logs you out and eats your your posting....



    Quote:

    The reason I wondered about that is that one often hears about how the Pentium translates the x86 instruction set into a more RISC-like instruction set. Since OS X doesn't have that old x86 legacy, couldn't it bypass that translation?



    It's too bad that 'experts' keep telling people this because it causes all kinds of confusion.



    The x86 architects did not suddenly wake up one morning and decide to put a translation layer around a RISC core and call it an x86. Just didn't happen; x86 is still CISC.



    More to the point, both RISC and CISC machines translate their instructions into 'macro-ops'. That process doesn't slow down the execution of the code. Everything is pipelined; It's all done on the fly.
  • Reply 16 of 49
    sillyfoolsillyfool Posts: 35member
    Quote:

    I am nearly sure that the internal code of a pentium M, have nothing to do with the one of a pentium 4.



    The jargon that you'll be hearing is 'micro-ops'. From the point of view of micro-ops, the internal architecture of the Pentium-M is a lot like that of the Pentium 4.



    The Pentium-M does something called 'micro-op fusion'. That improves the amount of work that it can complete in a clock cycle. So that helps to speed things up and reduce power.



    The improvements in the Pentium-M come from details of how the processing units are designed and how they are managed.



    Quote:

    So it is actually a different class of processors, isn't it?



    Well, not 'class' so much as generation. All of the Intel designs are heading down this path.



    Quote:

    Why the name "Pentium"?



    A side effect of copywrite law. Intel used to market their products using a number based system 286, 386, etc. They would copywrite the number '386' and that caused all kinds of problems. Eventually this was challenged in court and it was ruled that numbers can not be copywrited. So they called the next product PENTIUM. PENT from Penta for five (like pentagram) and IUM on the end just because they liked the sound of it.
  • Reply 17 of 49
    placeboplacebo Posts: 5,767member
    Wait, so is the Pentium-M another processor? I thought it was a notebook processor (where the M stood for mobile) for some reason.
  • Reply 18 of 49
    ruudruud Posts: 20member
    Quote:

    Originally posted by Placebo

    Wait, so is the Pentium-M another processor? I thought it was a notebook processor (where the M stood for mobile) for some reason.



    The Pentium M is a direct descendant of the Pentium 3 architecture. Up until now it's primarily used in notebooks (under the name Centrino). Basically it's a smart and efficient design, as opposed to the raw (but inefficient) horsepower of the P4. However, since it turns out that the Pentium 4 architecture (Netburst) is not as scalable as originally thought, new desktop processors are also going to be based on the Pentium M.
  • Reply 19 of 49
    brussellbrussell Posts: 9,812member
    Thanks THT and sillyfool and PowerDoc. And PowerDoc, you're going to have to change your username to PentiumDoc.
  • Reply 20 of 49
    placeboplacebo Posts: 5,767member
    Quote:

    Originally posted by ruud

    The Pentium M is a direct descendant of the Pentium 3 architecture. Up until now it's primarily used in notebooks (under the name Centrino). Basically it's a smart and efficient design, as opposed to the raw (but inefficient) horsepower of the P4. However, since it turns out that the Pentium 4 architecture (Netburst) is not as scalable as originally thought, new desktop processors are also going to be based on the Pentium M.



    Ah, I remember that now. Intel was forced to get better power consumption, so they looked at the chip and tweaked it from the ground up.
Sign In or Register to comment.