Okay, benchmarking geeks, here’s a word problem for you: Apple’s iMac with a 2.0 GHz Intel Core Duo is twice as fast as the old iMac with a 2.1 GHz G5. How does it compare to a Power Mac? Hard to say, especially since only some apps are native on the new chip, but here’s a sobering stat: in terms of Macworld.com’s Speedmark test, the Quad-Core 2.5GHz Power Mac G5 only bests a 2.1 GHz G5 iMac by 257 to 190. That means, running native code (soon to include Logic Pro), the iMac could perform as well or significantly better than a more expensive Power Mac. Don’t rush out and buy the iMac just yet, but can anyone explain?


Intel in Macs [Apple.com]


Okay, we’ve already got one reader who seems to think that true quad optimization will boost performance of the Power Mac. But that’s part of the point: a lot of Mac apps aren’t optimized for Velocity Engine or multi-processor, whereas compiling a universal binary — as is happening with the Apple apps — will be significantly easier. I think that will give the iMacs an edge on those applications, but I’m honestly not sure: way too many variables here. There’s a reason this story is posed as a question. Other folks want to chime in? -PK

  • erasmus

    apparently you didn't read all the way to the bottom of the article that you yourself linked in this post, for there is the answer to your own question – it clearly says that the benchmark tests are NOT NECESSARILY the best measure of what the Quad can do. Boot time & some of Apple's iLife apps are not optimized for or do not use the velocity engine/multiple processors. apps that do use it see a significant increase in speed.

  • admin

    . . . and it doesn't even begin to answer my question. I think developers will find it easier to go universal binary than add Velocity Engine or quad enhancements — most didn't do the latter, in fact, and now with Apple going Intel there's even less reason to do things for the G5. On the other hand, something like the AltiVerb reverb or parts of Peak that are Velocity Engine-enhanced should run much faster on the G5, and the iMacs have other bottlenecks (like graphics card). So this is definitely a very complex comparison. I just raise the question. That Speedmark stat isn't meaningless; it will be reflected in some apps — but not others. I'm just guessing that Logic Pro, for instance, will be happier on Intel in the short term. Apple wants those Power Mac Intels out as fast as it can, because then this all gets easier.

  • kokorozashi

    I'll babble on about some issues and you can decide which ones seem most relevant.

    [1] The compilers used for comparison are different. The PowerPC iMac runs code compiled with gcc and the Intel iMac runs code compiled with an Intel compiler. I don't know if or when this Intel compiler will be made available to developers. This is a big deal.

    [2] The CPU isn't the only performance factor. You've got memory buses and such to worry about.

    [3] Logic probably is optimized for PowerPC. It's been an Apple product for years. You'd think they would have put in AltiVec optimizations by now. I don't know, but it stands to reason. There is of course a sort of subset of AltiVec on Intel and presumably Apple would want to use that as well.

    [4] Intel processors are better at running code which has not been optimized because they in effect recompile code into smaller operations as the code runs, and that process involves some optimization. PowerPC processors are dependent on the compiler to do that kind of work, so if your PowerPC compiler sucks, that's just the way it is.

    [5] The iMac is a lot cheaper, has a monitor, and is cute. This is pretty important because in truth anybody who's just gotten around to maxing out a G5 iMac today stands to benefit tremendously by grabbing an Intel iMac. Those of us iMac fans with machines as "slow" as a 1GHz G4 are drooling profusely.

  • admin

    Hi kokorozashi — think you hit the nails on the head.

    I knew about the gcc / Intel compiler issue, but does that mean Apple's benchmark is completely bogus? Any indication of what this means for their own code / developers using Xcode?

    I will say, anecdotally, developers using the Intel test machines sent to some developers were impressed by the speed using the existing OS X developer tools. That's, of course, completely unscientific. (Not that Apple is entirely scientific.)

    As for optimizations in Logic Pro, I think that's probably anyone's guess. In a Macworld past, we were watching an unscientific demo that showed Logic Pro blazing on a G5, which I later determined (by grilling people at Apple) was a series of stacked reverbs. So it's possible those reverbs were optimized. Honestly, though, I've talked to many developers who are losing Velocity Engine enhancements, and they don't seem to think it's a big deal — at least, that it won't be a step backwards.

    So, in other words, everthing is clear as mud in terms of real performance. Guess I'll have to wait for more data, talk to more developers.

    But I think you outlined the main issues. And the important thing is, YES . . . just as upgrading from 1G G4 or less to the *previous* current generation felt like a big jump, anyone going for a new Intel iMac is likely to be very happy.

    Speaking as a cross-platform user, I'd also like to benchmark the performance hit taken on Windows by hitting your head against walls. (Hey, I love my PC, but . . . well, I ought to be able to graph that.)

    Peter

  • kokorozashi

    I don't know how to quanitfy the importance of compilers. I am not a compiler guy and you've pretty much already heard everything interesting I might have to say on the topic. :-)

    As far as hitting your head against walls while running Windows, you really need to stop flirting with those non-Reason applications. :-)

  • kokorozashi

    "In short, Apple used multiprocessor benchmarks to skew the performance advantage that its Intel-based machines enjoy compared to single-core PowerPC G4 and G5."
    http://weblog.infoworld.com/enterprisemac/archive

  • kokorozashi

    That said, it should be noted that well-architected music apps are just the sorts of apps which should benefit most from multi-core processors.

  • admin

    Thanks for those answers and that link! I knew if I raised this issue I'd get some educated answers. This also gives me some good questions to ask developers at NAMM.

    I would expect them to "skew" their benchmarks, as they want to show off multi-core and its unique benefits.

    One point I haven't heard made: I think we're in a very good place in computing evolution. There's an incredible amount you can do with a current-generation non-Intel Mac (or non dual-core PC). That doesn't mean things are slowing down, it means we're reaching maturity where someone who just bought a computer doesn't have fundamentally different capabilities than someone who bought one a year or two ago. That creates an unhealthy divide in computing that doesn't really benefit anyone. I'm convinced from what I've seen that we will see significant performance gains on the dual-core machines, and that some of those benefits will appear over time as software catches up. It's also not insignificant for cross-platform software (which includes a LOT of music software) that you'll have the same benefit on both platforms, not something time-intensive and limited in benefits like Velocity Engine has been on Mac.