New Mac performance claims

Posted:
in Current Mac Hardware edited January 2014
In what way has it been determined that the new MacIntels are multiple times faster than the PPC cpus that they replace?



The iMac is supposed to be twice as fast. Is that because it has two cores?



And the PowerBook(whoopsie), the Pro Mac Book, is supposed to be up to four times as fast as the G4 PowerBook. How are these things determined?



It will be wonderful if these things are true. I suppose Apple figures that the Yonah is like a G5, and there are two cores in the Yonah, and since the G5 is faster than the G4, they arrive at this.



Steve Jobs must have been frustrated with the PPC makers, to remark on how Power is gone.

Comments

  • Reply 1 of 19
    hmurchisonhmurchison Posts: 12,128member
    Judging from the keynote I think we can only say that the numbers come from SPEC_int and SPEC_fp scores.



    I've yet to read a report about the realworld speeds based on MWSF demo models. Bummer
  • Reply 2 of 19
    lundylundy Posts: 4,466member
    Go to http:///www.apple.com and click on Hardware. Then click the product that you want to read about. All the graphs and explanations are there.
  • Reply 3 of 19
    I read those pages yesterday, but not the SPEC's site, so I went there. But thanks for the reference anyway.



    You know, I do have a fair amount of trust in Apple's judgment about what Apple is doing.



    Now about their performance claims, I always believed every word that Apple wrote about the speed of its processors too, comparing PPC to x86. Did not everyone?



    The SPEC, and their plans in the future for measuring performance, as they are posting, those things I will read more about as time goes on.



    It will be interesting to read other evaluations of the speed of the Yonah compared to the G5, or the G4, in application tests.
  • Reply 4 of 19
    Quote:

    Originally posted by NordicMan

    Now about their performance claims, I always believed every word that Apple wrote about the speed of its processors too, comparing PPC to x86. Did not everyone?



    And now Apple is claiming the new Intel processors leave the current G5 and G4 PPC chips in the dust.



    The irony is killing me...
  • Reply 5 of 19
    You can never trust any company's own numbers wholesale. It's strictly PR.



    Apple's new performance numbers, for example, are under optimal conditions that arent likely to be repeated. "Universal" apps won't use Intel compilers, although that's what they used for their "tests."



    The new chips are still pretty fasy, regardless. But wait for reliable tech sites to run their own apples-to-apples tests, like Anandtech.com Eventually they'll have something.
  • Reply 6 of 19
    Apples to apples tests, that is good. Yes, that is what we wait for, is more tests. I love Apple's goods, I am not knocking too much, and I know that the announcements say 'up to', and are for pr. It just seemed a bit much. But intel and Apple should prove a good match. Often when you read comparisons of AMD and intel processors, the AMD are better for gaming, and for some apps, and the intel are a little better for some graphics or for some 3D applications. So it should be a good match, intel and Apple, for desktop, or workstation use. For the high end server and workstation, it seems that Opterons out perform any intel chips.



    I did not think that I would be for getting a new Mac right off(although I was hoping that new 13.3" book would be coming, as I have never owned nor used a Mac notebook). But now I am getting a hankering for one or the other of the two, iMac or Mac book. I am sure they are speedy. Exciting times.
  • Reply 7 of 19
    Quote:

    Originally posted by nowayout11

    "Universal" apps won't use Intel compilers, although that's what they used for their "tests."





    Huh, i was under the impression that xcode is able to generate universal binaries using intel compilers?
  • Reply 8 of 19
    placeboplacebo Posts: 5,767member
    Intel was developing a compiler to work with Xcode.



    "The new Intel compilers will act as plug-ins to Apple's Xcode development environment."



    -eWeek article
  • Reply 9 of 19
    Quote:

    Originally posted by Performa636CD

    And now Apple is claiming the new Intel processors leave the current G5 and G4 PPC chips in the dust.



    The irony is killing me...




    But that is how it works unfortunately, from a PR point of view. Its not that bad for the G5 actually, because it has such a nice FSB. The iMac is 2-3x times faster, and thats with dual-core, so 1-1.5x per core is, well, almost 1:1.



    The MacBook is creaming the PB though, but we all would have expectet that since its FSB is so much better.
  • Reply 10 of 19
    www.barefeats.com



    Stay tuned.
  • Reply 11 of 19
    dfilerdfiler Posts: 3,420member
    While watching the keynote with a friend, I made some snide remarks about misleading benchmarks.



    He thought it stemmed from some kind of PPC patriotism, that I disliked the new Macs. But this wasn't the case at all. I like the intel based Macs and am simply disappointed in Apple's deceptive advertising. Not that apple is unusual in this regard. Only, that in the past, they didn't stoop to using obviously artificial and purposefully misleading benchmarks.



    Remember when apple mocked others for using artificial benchmarks and how apple used real-world tests? (granted, mostly altivec optimized photoshop).



    Edit: The benchmarks steve quoted are great for chip designers talking about specific performance areas. They're "artificial" when quoted to the public as some kind of general benchmark.
  • Reply 12 of 19
    A good test/benchmark....



    Time to render a DV file to h.264 ... compare how long it takes to do a 30 minute video or some such thing.



    This would give a fairly realistic comparison... it uses CPU, lots of RAM access and lots of Disc access... would compare the actual machines as an entire unit rather than just raw CPU power.
  • Reply 13 of 19
    telomartelomar Posts: 1,804member
    Quote:

    Originally posted by Performa636CD

    And now Apple is claiming the new Intel processors leave the current G5 and G4 PPC chips in the dust.



    The irony is killing me...




    Keep in mind most reviews show Yonah outperforms Intel's old Netburst architecture comfortable, which is what Apple used to compare to. Add to that the tests they used are MP aware and you end up with around the expected improvements. I am a bit skeptical on FP performance but that might come down to latencies.
  • Reply 14 of 19
    Quote:

    Originally posted by dfiler

    While watching the keynote with a friend, I made some snide remarks about misleading benchmarks.



    He thought it stemmed from some kind of PPC patriotism, that I disliked the new Macs. But this wasn't the case at all. I like the intel based Macs and am simply disappointed in Apple's deceptive advertising. Not that apple is unusual in this regard. Only, that in the past, they didn't stoop to using obviously artificial and purposefully misleading benchmarks.



    Remember when apple mocked others for using artificial benchmarks and how apple used real-world tests? (granted, mostly altivec optimized photoshop).



    Edit: The benchmarks steve quoted are great for chip designers talking about specific performance areas. They're "artificial" when quoted to the public as some kind of general benchmark.




    The tests aren't artificial/deceptive at all. For starters, take the MacBook. While the actual SPEC tests are 4.5 - 5.2 times faster, Apple only advertises 4x faster, which is reasonable. You'll also notice they benchmarked Modo, which was about 4.1x faster. I'm assuming Modo uses the Intel compiler, although I'm not sure.



    Programs that are compiled in GCC are only about 2x faster- FCP, Safari, Doom, etc... However, GCC is notoriously unoptimized for Intel hardware. Intel specifically states that all floating point operations, vector or not, should be performed on the vector unit, SSE3. Intel's compiler does this. GCC does not?it compiles float operations to the FPU, which, as far as Intel is considered, is obsolete.



    Intel's compiler, which is coming out soon for Xcode, is C/C++ only. So, Carbon apps can take advantage of it. Cocoa apps cannot.



    This problem is temporary. Either Intel can add support for Objective-C to their compiler, or, more likely, Apple can patch GCC to properly compile x86 float ops as SSE instructions. This isn't really that difficult, it'd just have to be extensively tested. I'd venture to say that it's likely this will happen before Apple's pro apps are released in March.



    Meaning, that a lot of apps really will be about 4 times faster. Just like Modo is already.
  • Reply 15 of 19
    Quote:

    Originally posted by gregmightdothat



    Meaning, that a lot of apps really will be about 4 times faster. Just like Modo is already.




    I think the modo app is simply multi-processor friendly (as shown by the results on a Quad G5.) I may be wrong but I think modo is a Cocoa app which would mean it's not being compiled with Intel's compiler.



    You'd say "well then the MacBook Pro should only yield a 2x gain for modo" and you'd be right but also remember that the old G4 PowerBooks were ridiculously bandwidth starved. modo needs as much bandwidth as possible...so it's clear bandwidth was the biggest problem on the G4 PowerBooks and modo could have run almost twice faster on it. This explains the 4x speed increase on the new 'Books.
  • Reply 16 of 19
    Quote:

    Originally posted by kim kap sol

    I think the modo app is simply multi-processor friendly (as shown by the results on a Quad G5.) I may be wrong but I think modo is a Cocoa app which would mean it's not being compiled with Intel's compiler.



    You'd say "well then the MacBook Pro should only yield a 2x gain for modo" and you'd be right but also remember that the old G4 PowerBooks were ridiculously bandwidth starved. modo needs as much bandwidth as possible...so it's clear bandwidth was the biggest problem on the G4 PowerBooks and modo could have run almost twice faster on it. This explains the 4x speed increase on the new 'Books.




    Modo is almost certainly NOT a Cocoa app, considering that (a) it's cross-platorm and (b) it uses it's own interface. Even if the interface were Cocoa, the guts wouldn't be, since they're already programmed in C++ for Windows.



    That said, you're right, there's lots of reasons Modo could be faster, but keep in mind that Final Cut Pro is multi-threaded, too, but is, I believe, at least partly Cocoa.



    Also, Modo isn't a particularly bandwidth heavy app?it's mainly video processing that demands lots of bandwidth. Video does a few calculations on a lot of pixels, whereas 3d rendering does a lot of calculations on a few pixels.
  • Reply 17 of 19
    dfilerdfiler Posts: 3,420member
    Quote:

    Originally posted by kim kap sol

    This explains the 4x speed increase on the new 'Books.



    I think you'll be disappointed if you're expecting a 4x increase. The intel Macs are great, just not 4x faster.



    Although I'm hoping to be pleasantly suprised.



    Either way, we will know for sure relatively soon, after reputable sites run real (end-user) benchmarks.
  • Reply 18 of 19
    Quote:

    Originally posted by gregmightdothat

    Modo is almost certainly NOT a Cocoa app, considering that (a) it's cross-platorm and (b) it uses it's own interface. Even if the interface were Cocoa, the guts wouldn't be, since they're already programmed in C++ for Windows.



    That said, you're right, there's lots of reasons Modo could be faster, but keep in mind that Final Cut Pro is multi-threaded, too, but is, I believe, at least partly Cocoa.



    Also, Modo isn't a particularly bandwidth heavy app?it's mainly video processing that demands lots of bandwidth. Video does a few calculations on a lot of pixels, whereas 3d rendering does a lot of calculations on a few pixels.




    Using it's own interface and being cross platform does not exclude it from being a Cocoa app. For examples download Skype and Adobe Lightroom.



    A quick Google search seems to indicate modo is Cocoa (although I must admit those that say this don't seem sure or understand. At first glance, you'd think that they believe an app is automatically Cocoa if it's developed in Xcode, which is not the case at all.)



    Ok...the guts could be programmed in C++...but if the interface is Cocoa, that would still make modo un-compilable using Intel's compiler.



    As for modo not being bandwidth heavy...I guess. But the bandwidth needed scales with the number of textures and the size of the textures used.
  • Reply 19 of 19
    hmurchisonhmurchison Posts: 12,128member
    Modo programmers said they have a base codebase which handles a majority of the functinality and then they add in platform specific code on top so if Modo is Cocoa it's likely to only be %25 Cocoa.
Sign In or Register to comment.