Since the first rumors of an
Apple switch to Intel, everyone has been wondering about the potential speed of Intel-based Macs. Last week’s announcement of the
first shipping Intel-based Macs brought with it the promise of a major speed boost: Apple’s
Web pages suggest that the new iMac, powered by the
Intel Core Duo processor, is twice as fast as its G5 predecessor.
Macworld Lab’s tests do show that the new Intel-based iMac is faster than the
iMac G5 when running native applications. However, we found that those improvements are generally much less than what Apple claims is a 2x improvement in speed.
Instead, our tests found the new 2.0GHz Core Duo iMac takes rougly 10 to 25 percent less time than the G5 iMac to perform the same native application tasks, albeit with some notable exceptions. (If you’d prefer, that makes the Core Duo iMac 1.1 to 1.3 times as fast.) And we also found that applications that aren’t yet Intel-native—which must run using Apple’s
Rosetta code-translation technology —tend to run
half as fast as the same applications running natively on the iMac G5.
Universal applications: Comparing apples to apples
Apple’s much-publicized test scores for the new iMac were made with programs designed specifically to generate test results. So they may give some indication of the overall performance potential of these systems. However, such test results often don’t match up to what regular users see in their everyday work—i.e., the speed of real-world applications.
iMac Core Duo
Several of the programs Macworld Lab uses to judge overall system performance are not yet available in Universal versions. As a result, we added several new tests based on applications that are currently available in Universal form. (And we plan on running even more tests as more Universal applications arrive—we’ll be posting test updates as well as a complete review of the new iMac to Macworld.com in the coming days.)
In tests with two
iLife ’06 applications—iMovie and
iPhoto —we found remarkably different performance depending on what features of the programs we tried. For example, the act of applying one iMovie effect to a video clip resulted in a remarkable speed improvement of 1.8 times. But a different effect showed only half the improvement, and yet another showed no speed improvement at all.
First Tests: Universal Applications
| ||iMac Core Duo/2GHz||iMac G5/2.1GHz||Times as fast
|iMovie 6.0 (B&W effect)||1:45||3:11||1.82x|
|iMovie 6.0 (Aged effect)||1:12||1:12||1x|
|iMovie 6.0 (Rain effect)||2:05||2:43||1.3x|
|iPhoto (export to QuickTime)||1:08||1:31||1.34x|
|iPhoto (export to Web)||2:01||2:12||1.09x|
|iPhoto (export to files)||2:55||2:40||.91x|
|Create Zip archive||2:32||2:53||1.14x|
Best results in bold.
All scores are in minutes:seconds. All systems were running Mac OS X 10.4.4 with 512MB of RAM, with G5’s processor performance set to Highest in the Energy Saver preference pane. Using iMovie, we applied 3 different video effects to a 1-minute movie, one at at a time. Next, we imported 100 JPEG photos into iPhoto and then exported them as a QuickTime movie, as a Web page, and as files, resized to be not more than 2,000-by-1,500 pixels. We created a Zip archive from a 1GB folder. We converted 45 minutes of AAC audio files to MP3 using iTunes’ High Quality setting. We saved an iDVD project containing a 6-minute, 46-second movie as a disk image. We used iSquint to compress the same movie for iPod video playback. We used BBEdit to run a Text Factory containing five editing, replacement, and sorting tasks on a 75.1 MB text file.—Macworld Lab testing by James Galbraith and Jerry Jung
Similarly, importing 100 photos into iPhoto 6 took 35 percent less time on the Intel-based iMac, and exporting from iPhoto to a QuickTime movie took 25 percent less time. But exporting iPhoto images to a Web page took only 8 percent less time. And exporting those images to files actually took 9 percent more time on the Intel-based Macs.
Bare Bones Software’s BBEdit performed well on the new systems. In our BBEdit test battery, which performed five operations (Process Lines Containing, two Replace All grep searches, Process Duplicate Lines, and Sort Lines) on a 75.1 MB text file, the Intel-based iMac took 21 percent less time than the G5.
Other tests—creating a Zip archive in the Finder, encoding an audio file in iTunes, and generating a DVD image (including all required MPEG-2 rendering) in
iDVD —resulted in the most common range of speed improvements, taking between 11 and 12 percent less time on the Intel-based iMac than on the G5.
These new iMacs are the first Mac systems to use the new Extensible Firmware Interface system for booting, and that may be one of the contributing factors to excellent start-up times for these systems. The Intel-based iMac booted in 25 seconds, almost half the time it took the iMac G5.
One application, however, constantly disappointed us during our testing: iMovie 6. Not only was this brand-new version of Apple’s video-editing application equally buggy on both platforms, but it was dramatically slower at compressing and exporting video on the Intel-based system than on the G5—so much so that we suspect iMovie’s poor performance is the result of a bug within iMovie rather than any intrinsic failure of the iMac.
Rosetta: Half-speed compatibility
One of the biggest mysteries about the Intel transition has been Rosetta, the technology that lets Intel-based Macs run programs compiled for Macs containing PowerPC processors. Just how fast will Rosetta run non-Universal programs? With this new iMac, we’re finally able to get the first clear answer to that question.
We tested three different non-native applications on the Intel-based iMac, and compared the results to the iMac G5. All of the tests showed that PowerPC applications run on the new iMac at less than half their native speed. Our standard Microsoft Word scrolling test ran at 48 percent of the speed it ran on the iMac G5; our standard battery of 14 Photoshop CS tasks ran at 45 percent of the G5’s speed; an MP3 encode using the non-native iTunes 6.0.1 ran at 34 percent of the speed.
Unlike most of Apple’s applications, BBEdit 8.2.4 offers an “Open using Rosetta” option that forces a Universal application to run in Rosetta. As a result, we were able to re-run our BBEdit test on the Intel-based iMac, but this time in Rosetta. The results were very much in line with our other tests: our BBEdit test took roughly twice as long to run within Rosetta as it did when allowed to run natively on the Intel processor.
First Tests: Applications in Rosetta
| ||iMac Core Duo/2GHz||iMac G5/2.1GHz||Times as fast as G5|
Best results in bold.
All scores are in minutes:seconds. All systems were running Mac OS X 10.4.4 with 512MB of RAM, with G5’s processor performance set to Highest in the Energy Saver preference pane. We converted 45 minutes of AAC audio files to MP3 using iTunes’ High Quality setting. We scrolled a 500-page document using Microsoft Word. The Photoshop Suite test is a set of 14 scripted tasks using a 50MB file. Photoshop’s memory was set to 70 percent and History was set to Minimum.—Macworld Lab testing by James Galbraith and Jerry Jung
The big question is, will applications running at half speed within Rosetta be fast enough for users? Many applications aren’t particularly processor-intensive, and will probably seem quite usable under Rosetta. (Microsoft Office, for example, will probably be fairly usable.) Other hard-working applications—Photoshop comes to mind—won’t fare as well. And using a non-native rendering application or game would seem to be a no-go.
The speed of Rosetta will also depend a lot on your perspective. Yes, if you’re upgrading to an Intel-based iMac from an iMac G5 you bought just a few months ago, all of your non-Universal software will run at half speed. But if you’re upgrading from a two- or three-year-old iMac, the speed difference may be much less noticeable.
The speed of applications running under Rosetta will be something to keep in mind, especially when it comes to the forthcoming release of the
MacBook Pro. The users of that professional-level laptop are far more likely to demand serious speed from their applications; if there’s no Universal version of Photoshop available at the time, professional photographers may balk at the idea of running Photoshop at a fraction of its speed.
What it means
If you talk to both Apple and Intel, they’ll tell you that the Intel Core Duo is a processor designed for laptops, providing a compromise between performance and good power-consumption and heat-generation characteristics. And so the Core Duo processor in these new iMacs (as well as the forthcoming MacBook Pro) is clearly not meant to be the be-all, end-all when it comes to raw computing power.
That’s one reason why Apple’s initial speed claims of doubled performance (with some tests running showing as much as a 3x speed boost) were so breathtaking, since they were coming from a chip meant to run small and cool. Unfortunately, our tests suggest that the remarkable results of Apple’s published tests aren’t reflected in most of the real-world applications we tested. Based on our initial tests, the new Core-Duo-based iMac seems to be 10 to 20 percent percent faster than its predecessor when it comes to native applications, with some select tasks showing improvement above and beyond that.
Potential iMac buyers who predominantly rely on applications that aren’t available in Universal versions (or, for that matter, those who rely on Classic, which is incompatible with Intel-based Macs) will likely not be interested in these first Intel systems. Running a handful of programs in Rosetta seems reasonable, but if you rely on numerous applications that aren’t yet Universal, it’s probably wise to wait.
After running these tests, several questions come to mind—and they impact the entire Mac product line, not just the iMac. How optimized is Mac OS X 10.4.4 for Intel processors, and will Apple be able to improve performance on these systems over time? Is there room for improvement in the optimization of individual applications, or is this all we’re going to get? How quickly will Intel improve processor performance from here, both in the Core Duo line and in other chip lines that will wend their way into future Apple products?
We don’t have all those answers yet, although several conversations with Apple representatives have suggested that developers will be able to improve the performance of their software as they become more comfortable with the new Intel architecture. Even Apple’s own freshly-released iLife ’06 applications could stand to be better optimized—and Apple representatives admitted as much to us.
How much speed can be wrung out of future improvements to software and the operating system itself? We simply don’t know. There’s reason for hope, but there’s much more testing to be done, many more Universal applications to be written, and more Mac OS X system updates to be released.
[ Jason Snell is the editorial director of Macworld . ]
(Updated on January 18 at 4:46 PM ET to add “times faster than” equivalents to charts. Updated again at 5:35 PM ET to attempt to clarify percentages slower v. percentages faster v. times as fast. Updated on January 20 at 1:43 PM ET to correct a line in the story text which switched the description of two iPhoto tests.)