New IBM Power PC 1GHz chip

2

Comments

  • Reply 21 of 60
    [quote]Originally posted by antaisce:

    <strong>





    Sorry, but this is BS. You should better have a look at the amount of RAM installed.

    </strong><hr></blockquote>





    Also TiBook has 1MB L2 cache while iMac has either 512 L2, or 256 (on chip).



    So you are not just comparing G4 or G3 there are other factors as well...
     0Likes 0Dislikes 0Informatives
  • Reply 22 of 60
    powerdocpowerdoc Posts: 8,123member
    [quote]Originally posted by Outsider:

    <strong>I would imagine IBM would make it Altivec compatible. They couldn't use the name Altivec (Apple doesn't even use it) maybe they will come up with their own name.</strong><hr></blockquote>

    If they want Apple for customers, this simd unit need to be altivec compatible, otherwise everything is permitted ... for the worse



     0Likes 0Dislikes 0Informatives
  • Reply 23 of 60
    msleemslee Posts: 143member
    The display layer of quartz is vector-based, so there is some benefit from having AltiVec there. For example, side-by-side comparisons of a G4 400 Powerbook and a G3 500 iBook indicated there?s a substantial difference between the two, as far as resizing live windows and strafing a magnified dock. Since the display layer calculations are not hardware accelerated, we can assume that the superior memory bandwidth of the Powerbook (faster bus, beefy L2 cache) is not coming into play, as most of the calculations are occurring in the CPU. So its stupid to say that OSX does not benefit from AltiVec, because it clearly does, and Apple itself has said as much. But make no mistake, AltiVec is no panacea for a low-clocked CPU. That 750Fx is going to haul some balls, no doubt. Granted, IBM has had to pull a lot of sleight-of-hand CMOS tricks to get the clock this high, Low-K dielectric, SOI...but the processor is a mean machine. The large L2 cache with its bigger pipe is going to be meaningful, as will the 200MHz bus. All in all, IBM estimates a 8-12% increase in performance, clock-for-clock, over the previous generation?s G3. At 1 GHz, this puppy is going to absolutely scream. And its low power consumption, and its ability to scale its power consumption on the fly, means that this G3 is a perfect match for the iBook. I hope to God Apple puts this thing in an iBook as soon as possible, because I would totally get one. This G3 is what IBM had in mind all along when they joined Apple and Motorola?no frills, no SIMD, just great design and excellent execution. There probably won?t be an SIMD on this processor, because it really isn?t in the IBM RISC philosophy. But don?t worry about that, because this processor will more than make up for any SIMD-optimizations in Quartz.
     0Likes 0Dislikes 0Informatives
  • Reply 24 of 60
    jcgjcg Posts: 777member
    [quote]Originally posted by discstickers:

    <strong>The orginal G4 was basically G3+Altivec. The G3 was based on the 603e core.



    I don't think its unreasonable that the iBook will get a G4 eventually. The Pismo and the orginal iBook had G3s. But I don't think there will be a revision until MWNY at the earliest.</strong><hr></blockquote>



    If I remember right the G4 is based off of the 604e core, thus it is capable of multi-processing. The G3 is based of of the 603e, which can not be set up as a MP system.
     0Likes 0Dislikes 0Informatives
  • Reply 25 of 60
    msleemslee Posts: 143member
    The newer G3's have resolved the cache-coherency problem that prevented a MP setup. IBM actually offers dual-G3 boards (developers).
     0Likes 0Dislikes 0Informatives
  • Reply 26 of 60
    powerdocpowerdoc Posts: 8,123member
    [quote]Originally posted by JCG:

    <strong>



    If I remember right the G4 is based off of the 604e core, thus it is capable of multi-processing. The G3 is based of of the 603e, which can not be set up as a MP system.</strong><hr></blockquote>

    The G4 (first version) is based upon the 603 core also, but he has also better new features, the altivec unit and the large pipeline that can feed him, a double precision fp unit 25 % more performant at equal mhz than the G3 and SMP capabilities.

    The 7450 line where based upon a new integer unit pipeline of 7 stages. (different architecture than the initial 603 core) and more independant altivec unit.

    No others chips was based upon the 604 core. The 604 was a very good chip for FP , but was far behind the G3 in the other areas, that's why this family of chip was discontinued.

    The G5 is supposed to be based upon a new core.



    [ 02-05-2002: Message edited by: powerdoc ]



    [ 02-05-2002: Message edited by: powerdoc ]</p>
     0Likes 0Dislikes 0Informatives
  • Reply 27 of 60
    [quote]Originally posted by powerdoc:

    <strong>

    The G4 (first version) is based upon the 603 core also, but he has also better new features, the altivec unit and the large pipeline that can feed him, a double precision fp unit 25 % more performant at equal mhz than the G3 and SMP capabilities.

    The 7450 line where based upon a new FP pipeline of 7 stages. (different architecture than the initial 603 core) and more independant altivec unit.

    No others chips was based upon the 604 core. The 604 was a very good chip for FP , but was far behind the G3 in the other areas, that's why this family of chip was discontinued.

    The G5 is supposed to be based upon a new core.



    [ 02-05-2002: Message edited by: powerdoc ]</strong><hr></blockquote>



    Yup, and thats why the G5 will kick every ass it can find.
     0Likes 0Dislikes 0Informatives
  • Reply 28 of 60
    outsideroutsider Posts: 6,008member
    powerdoc is right. The 604 core ended with the 604ev. G4 is based off the G3 which in turn is based off the 603.
     0Likes 0Dislikes 0Informatives
  • Reply 29 of 60
    amorphamorph Posts: 7,112member
    [quote]Originally posted by JCG:

    <strong>



    If I remember right the G4 is based off of the 604e core, thus it is capable of multi-processing. The G3 is based of of the 603e, which can not be set up as a MP system.</strong><hr></blockquote>



    The G4 (7400 and 7410) are also 603e-derived, although the FPU was adapted from the 604e.



    The original G4s had better SMP support than the 604e.



    There was a companion to the 750 G3 that was derived from the 604e, and there was a high-end motherboard that went with it, but both were scrapped as needlessly expensive when Apple found out just how fast the G3 was. That's why their high-end machines were shipping with a board named "Gossamer" rather than, say, "Tsumani" - it wasn't intended to be their high-end offering.
     0Likes 0Dislikes 0Informatives
  • Reply 30 of 60
    vmxvmx Posts: 10member
    [quote]Originally posted by powerdoc:

    <strong>

    The G4 (first version) is based upon the 603 core also, but he has also better new features, the altivec unit and the large pipeline that can feed him, a double precision fp unit 25 % more performant at equal mhz than the G3 and SMP capabilities.

    The 7450 line where based upon a new integer unit pipeline of 7 stages. (different architecture than the initial 603 core) and more independant altivec unit.

    </strong><hr></blockquote>

    I believe that it would be more precise to say that the G4 is based on the G3, not the 603. Yes, the 750 was based on the 603 family, so indirectly the 7400 owes its lifeblood to the 603, but you may as well say that the 7400 was derived from the 601 at that point. 601 -&gt; 602 -&gt; 603 -&gt; 750 -&gt; 7400 (from the 10k foot view).
     0Likes 0Dislikes 0Informatives
  • Reply 31 of 60
    [quote]Originally posted by discstickers:

    <strong>The orginal G4 was basically G3+Altivec.

    </strong><hr></blockquote>



    The G4 can also do single-cycle double-precision floating point operations, and I'm not sure the data stream instructions are available on the G3 either. In addition, the G4 supports MPX bus, which is much more efficient than 60x.



    Bye,

    RazzFazz



    [ 02-05-2002: Message edited by: RazzFazz ]</p>
     0Likes 0Dislikes 0Informatives
  • Reply 32 of 60
    [quote]Originally posted by BRussell:

    <strong>I know it's conventional wisdom that Altivec speeds up OS X, but I just don't buy it. iMovie, Photoshop, iDVD, sure. But not menu selecting, window resizing, disc and network access, and the other things the OS does.



    Altivec is good for processing long continuous streams of data like occur in encoding or transforming multimedia, not the mostly short bursts of discrete events that occur in the OS.

    </strong><hr></blockquote>



    AltiVec doesn't need any large streams to work, just anything that applies the same operation to 4 - 16 elements at a time. This is true for 2D graphics, including transforms, transparency, and others.



    Bye,

    RazzFazz
     0Likes 0Dislikes 0Informatives
  • Reply 33 of 60
    programmerprogrammer Posts: 3,503member
    AltiVec is also really outstanding at reordering & moving data -- things which the OS' network layer does a lot of. I seem to recall the Apple networking people reporting that they could use AltiVec to accelerate quite a few system functions. Who knows if they've actually gotten around to doing that or not, however.
     0Likes 0Dislikes 0Informatives
  • Reply 34 of 60
    stimulistimuli Posts: 564member
    To add to Outsider's post above, the 750FX will have:



    Higher core frequency

    512KB on chip L2 cache at core frequency

    on a 256Kb bus

    lots of nifty little enhancements

    faster bus support

    consecutive read operations

    miss-under-miss cache lookups

    0.13 micron process

    lo-k dieelectric

    SOI

    34.6 mm² (That's less than 6x6mm!)

    3.6W @ 800 MHz

    5 stage pipeline (as opposed to the current G4's 7, meaning reduced latency!)



    Make no mistake, this is one mean little G3. I too would take a 1Ghz G3 over an 800mhz G4, as Integer ops are a hell of a lot more common than you RDF-ed people seem to think. &lt;b&gt;Especially&lt;/b&gt; in GUI.





    Together, the cache and bus improvements can increase system performance more than 25% over the 750CXe, when operating at the same bus frequency in bus intensive applications.



    [ 02-05-2002: Message edited by: stimuli ]</p>
     0Likes 0Dislikes 0Informatives
  • Reply 35 of 60
    powerdocpowerdoc Posts: 8,123member
    [quote]Originally posted by stimuli:

    <strong>To add to Outsider's post above, the 750FX will have:



    Higher core frequency

    512KB on chip L2 cache at core frequency

    on a 256Kb bus





    [ 02-05-2002: Message edited by: stimuli ]</strong><hr></blockquote>



    Wow256 Kb data bus : absolutely terrific this stuff has a memory bandwitch of 30 terabyte per second at 1 ghz. <img src="graemlins/lol.gif" border="0" alt="[Laughing]" />



    [ 02-05-2002: Message edited by: powerdoc ]</p>
     0Likes 0Dislikes 0Informatives
  • Reply 36 of 60
    [quote]Originally posted by stimuli:

    <strong>...Integer ops are a hell of a lot more common than you RDF-ed people seem to think. &lt;b&gt;Especially&lt;/b&gt; in GUI. </strong><hr></blockquote>



    Heh, you haven't looked at the technical details of AltiVec, have you? Most of AltiVec's instructions operate on integers... but 16, 8, or 4 of them at a time depending on whether you are working on 8, 16, or 32 bits integers. It also includes instructions that are idea for 2D graphics that the integer units don't have -- add w/ saturation, for example. The AltiVec unit is ideal for doing what takes time in a GUI: drawing pixels.
     0Likes 0Dislikes 0Informatives
  • Reply 37 of 60
    stimulistimuli Posts: 564member
    Actually, powerdoc, that's 256 Kilobits, not Bytes. So your looking at (theoretically) 30+ GB (Bytes)/sec throughput at 1 Ghz. Nothing to sneeze at.



    Programmer, is that the latest G4s, or also the 7400/7410?
     0Likes 0Dislikes 0Informatives
  • Reply 38 of 60
    amorphamorph Posts: 7,112member
    [quote]Originally posted by stimuli:

    <strong>Programmer, is that the latest G4s, or also the 7400/7410?</strong><hr></blockquote>



    That's true of all of the G4s.
     0Likes 0Dislikes 0Informatives
  • Reply 39 of 60
    stimulistimuli Posts: 564member
    Interesting, because Macs tend to suck at Integer ops compared to beasts like say, the Athlon.
     0Likes 0Dislikes 0Informatives
  • Reply 40 of 60
    eugeneeugene Posts: 8,254member
    stimuli, what do you think RC5 is? RC5 is pure integer. If only people would take the effort to write AltiVec code into their apps...
     0Likes 0Dislikes 0Informatives
Sign In or Register to comment.