Since the publication of Macworld’s
first lab tests of the Intel-based iMac and our subsequent
review, there’s been a massive amount of discussion about our results. There’s also been a lot of confusion and plenty of heated discussion. All the while, we’ve continued to test the new Intel macs, including adding some new tests as new
Universal applications arrive on the scene.
So as the
Intel transition moves into the second month of 2006, here’s an update on what we’ve learned so far about how Intel-based Macs perform.
2x or not 2x?
When Steve Jobs stood on stage at Macworld Expo in January, he claimed that the new iMac ran as much as two to three times faster than the previous iMac, and the new MacBook Pro laptop ran as much as four to five times faster than the PowerBook G4. There was loud applause from the crowd and a sigh of relief from Mac-watchers everywhere.
Of course, as Jobs himself pointed out, those performance claims were based on some very specific tests. When Macworld tested the new Intel-based iMac, we found that speeds varied widely—but that many tasks commonly run by Mac users today only run between 1.1 and 1.3 times as fast (or, if you prefer, 10 to 30 percent faster) on the Intel-based 2GHz iMac as on the 2.1GHz iMac G5. So while Apple’s iMac web pages prominently display a badge proclaiming “2x faster,” the reality is far more complicated—and for most typical uses, far less dramatic.
But let’s be clear: Apple wasn’t fibbing when it reported that certain tests showed that the new Intel iMac was faster than its predecessor by factors from 1.9 to 3.2. However, those tests were carefully selected to show the new iMacs at their very best, ideal cases designed to put the the new Intel Core Duo chip that powers these systems in the best possible light.
We did see similarly dramatic results in a few of our tests. Rendering a 3-D scene using the Java-based program
Art of Illusion, the Intel-based iMac was 3.2 times as fast as its G5 counterpart. (It was also faster than a dual-processor Power Mac G5, suggesting that these Intel Macs are much better than PowerPC-based models when it comes to Java performance.) In two tests using a not-yet-public Universal beta of Aspyr’s Doom 3, the new iMac was 1.7 and 2.1 times better. Yet many of our other tests showed comparatively modest speed improvements, with the Intel-based system commonly between 1.1 and 1.3 times as fast as the G5 model.
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. More strangely, both of our attempts to export from iMovie (to an iPod and to a Web-encoded movie) were slower on the Intel-based Mac. (Apple says it’s investigating the issue.)
Importing 100 photos into iPhoto 6 was 1.5 times faster, and exporting from iPhoto to a QuickTime movie was 1.3 times faster. But exporting iPhoto images to a Web page was only slightly faster. And exporting those images to files was actually slower on the Intel-based Macs.
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, between 1.1 and 1.2 times as fast on the Intel-based iMac as on the iMac G5.
Universal application tests
| ||iMac Core Duo 2GHz ||iMac Core Duo 1.83GHz ||iMac G5 2.1GHz ||X Factor |
|Startup || 0:25 || 0:25 || 0:46 ||1.8 |
|iMovie 6.0 (export to iPod) ||6:39 ||7:14 || 6:00 ||0.9 |
|iMovie 6.0 (export to Web) ||9:20 ||10:20 || 3:20 ||0.4 |
|iMovie 6.0 (B&W effect) || 1:45 ||1:54 || 3:11 ||1.8 |
|iMovie 6.0 (Aged effect) || 1:12 ||1:19 || 1:12 ||1 |
|iMovie 6.0 (Rain effect) || 2:05 ||2:12 || 2:43 ||1.3 |
|iPhoto (import 100 JPEGs) || 0:53 ||0:55 || 1:22 ||1.5 |
|iPhoto (export QuickTime movie) || 1:08 ||1:12 || 1:31 ||1.3 |
|iPhoto (export Web site) || 2:01 ||2:12 || 2:12 ||1.1 |
|iPhoto (export files) ||2:55 ||3:06 || 2:40 ||0.9 |
|Zip archive || 2:32 ||2:44 || 2:53 ||1.1 |
|Doom 3 || 35.9 ||27.3 || 17.0 ||2.1 |
|iTunes 6.0.2 || 1:14 ||1:20 || 1:23 ||1.1 |
|Art Of Illusion 2.2 || 1:55 ||2:03 || 6:07 ||3.2 |
|iDVD 6 || 13:55 ||14:47 || 16:25 ||1.2 |
|BBEdit || 1:37 ||1:41 || 2:02 ||1.3 |
Best results in bold; reference system in italic. X Factor refers to the number of times faster the iMac Core Duo 2GHz is versus the iMac G5 2.1GHz (1.0 = identical speed).
All scores are in minutes:seconds except for Doom results, which are in frames per second. 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 compressed a 6-minute, 46-second movie for iPod transfer and for the Web using default settings. We also applied 3 different video effects to a 1-minute movie, one at at a time. We imported 100 JPEG photos into iPhoto and then exported them as a QuickTime movie, a Web page, and as files resized to be not more than 2,000-by-1,500 pixels. We created a Zip archive in the Finder from a 1GB folder. We ran the timedemo 1 test on a beta Universal version of Doom 3 version 1.3.1303 set to use Ultrahigh graphics at 1024×768 resolution with all advanced options set to Yes except for vertical sync and antialiasing. We converted 45 minutes of AAC audio files to MP3 using iTunes’ High Quality setting. We rendered a scene using the Java-based Art Of Illusion 3-D program. We saved an iDVD project containing a 6-minute, 46-second movie as a disk image. 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
Are Two Cores Better than One?
There are numerous reasons for such variations in test results. But by far the most important has to do with the dual-core nature of the Intel Core Duo.
In the past, Macs got faster largely because the clock speeds of the processors got faster—for example, a 1.2GHz PowerBook G4 was clearly faster than a 1GHz PowerBook G4. However, there’s another way to make a Mac faster: add more processors. Apple’s used this approach before. Multiprocessor Power Macs have been available for a while; all the current Power Mac G5 models use dual-core chips, which essentially contain two processors on one physical chip.
Here’s the catch, though: adding processors to a Mac doesn’t automatically boost system speed the way increasing the clock speed does. That’s because programs must be specifically designed to support multiple processors to gain any benefit.
Since powerhouse Power Macs have supported multiple processors for years now, many heavy-duty professional programs—including graphics tools such as Adobe Photoshop, 3-D tools such as Maya and Cinema 4D, and video tools such as Final Cut Pro Studio and Adobe After Effects—have been modified to take advantage of multiprocessing. As Intel-specific versions of such apps arrive, they’ll undoubtedly perform well on these new iMacs.
But these iMacs are consumer systems. And many consumer-level apps don’t really take advantage of multiprocessing. To really take advantage of the Core Duo’s second processor, such programs will need to be updated to add better multiprocessor optimization. (Note that, if you’re running several programs at once, Mac OS X is smart enough to spread them out across multiple processors. That can provide a speed boost if a user is multitasking, switching between several processor-intensive programs at once.)
To find out just how much our test programs took advantage of the iMac’s dual-core chip, we disabled one of the iMac’s two cores and re-ran several of our tests. The results showed, for example, that some tasks in iPhoto (importing images and exporting a QuickTime movie) took advantage of the second processor core much more than others (exporting to files and Web pages). iTunes is very good at using both processor cores for ripping MP3s, while the Finder seems to only use a single processor when creating our Zip archive.
Testing Multiprocessor Performance
| ||iMac Core Duo 2GHz || iMac Core Duo 2GHz
(1 processor) || “Processor
|iMovie 6.0 (B&W) ||1:45 ||2:18 || 1.3 |
|iMovie 6.0 (Aged) ||1:12 ||1:43 || 1.4 |
|iMovie 6.0 (Rain) ||2:05 ||2:41 || 1.3 |
|iPhoto import 100 files ||0:53 ||1:29 || 1.7 |
|iphoto to QT Movie ||1:08 ||1:36 || 1.4 |
|iPhoto to Web page ||2:01 ||2:13 || 1.1 |
|iPhoto to file (2k x 1.5k) ||2:55 ||2:59 || 1.0 |
|Zip Archive ||2:32 ||2:33 || 1.0 |
|Doom 3 v1.3.1303 Universal (Beta) demo1 ||35.9 ||34.3 || 1.0 |
|iTunes 6.0.2 ||1:14 ||2:04 || 1.7 |
|Art Of Illusion render ||1:55 ||3:37 || 1.9 |
|iDVD 6 ||13:55 ||18:39 || 1.3 |
| iTunes 6.0.1 ||3:41 ||6:26 || 1.7 |
| Photoshop CS2 ||2:50 ||3:35 || 1.3 |
| Word Scroll ||1:58 ||2:03 || 1.0 |
Rosetta applications in italic. Processor Factor refers to the number of times faster the test ran with both processor cores enabled (1.0 = identical speed).
All scores are in minutes:seconds, except for Doom results, which are in frames per second. 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 used the same settings as in our standard tests, but used Apple’s CHUD performance tools to disable the second processor core and re-run the tests.—Macworld Lab testing by James Galbraith and Jerry Jung
Rosetta: Compatibility with a Caveat
There’s yet another wrinkle in the performance of these new Intel-based iMacs: Different processors speak different languages. Programs compiled for the PowerPC chip can’t run natively on these new Intel CPUs. Instead, they have to run through Rosetta, an emulation technology that lets Intel-based Macs run PowerPC apps by translating their commands into ones Intel chips can understand.
All of our tests showed that PowerPC applications run less than half as fast on a 2GHz Intel iMac than on the 2.1GHz iMac G5. But those speed differences are relative. Yes, if you’re upgrading to an Intel-based iMac from an iMac G5 you bought just a few months ago, your PowerPC apps will run half as fast. But if you’re upgrading from a two- to four-year-old iMac, you might not see any slow-down—and you might see a speed-up.
For example, when we compared the new 2GHz iMac to a 700MHz iMac G4, the first-generation flat-panel iMac released in 2002, we found that the Intel iMac executed our Rosetta application tests much faster than its four-year-old predecessor.
Since many of the everyday applications people use aren’t particularly processor-intensive (Microsoft Office being a great example) those applications should probably seem quite usable under Rosetta. Other programs (games, for instance) may be well-nigh unusable.
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. However, given how relatively slow the current PowerBook G4 line may be, the situation may not be that bad. And since Photoshop can take advantage of the MacBook Pro’s dual-core processor, even a non-Intel version of Photoshop may be quite usable on the MacBook Pro compared to current PowerBook G4s. But we won’t know for sure until the MacBook Pro arrives and we have a chance to put it through the same tests as the Core Duo iMac.
iMac: Rosetta Application Tests
| ||iMac Core Duo 2GHz ||iMac Core Duo 1.83GHz || iMac G5 2.1GHz || eMac 1.42 GHz G4 || iMac 700MHz G4 |
|iTunes 6.0.1 ||3:41 ||3:59 || 1:16 || 2:15 || 5:08 |
|Photoshop CS2 ||2:50 ||3:02 || 1:16 || 1:49 || 3:33 |
|Word Scroll ||1:58 ||2:05 || 0:57 || 1:29 || 3:07 |
Best results in bold; reference systems in italic.
All scores are in minutes:seconds. All systems were running Mac OS X 10.4.4 with 512MB of RAM, except the G4 iMac which was tested with 640MB of RAM. The G5’s processor performance was set to Highest in the Energy Saver preference pane. We converted 45 minutes of AAC audio files to MP3 using iTunes’ High Quality setting. The Photoshop Suite test is a set of 14 scripted tasks using a 50MB file. Photoshop’s memory was set to 70 percent, except for the G4 iMac which was set to 56%, and History was set to Minimum. We used Microsoft Word to scroll a 500 page document.—Macworld Lab testing by James Galbraith and Jerry Jung
When Universal Isn’t Enough
But just because a program is Universal doesn’t mean that it’s reached its full speed potential on Intel-based Macs. Software development is an evolutionary process. Even if software vendors (including Apple) have delivered Universal versions of their apps in time to run on these first Intel iMacs, there’s every reason to believe they’ll continue to tweak and refine those apps to work better and better on the new CPUs. For example, Apple officials have admitted to us that iLife ‘06 could stand to be better optimized.
Those developers have already gone through that optimization process for PowerPC-based Macs. Some programs, for example, offer features written to specifically take advantage of the G4 and G5 processors’ Velocity Engine, a special set of accelerated commands. Many of these features will need to be modified to take advantage SSE/MMX, the Intel equivalent of Velocity Engine.
In addition, the tools that developers use—Apple’s Xcode development environment and Intel’s collection of code compilers—will also likely improve, making it easier for developers to wring even more performance out of their Intel-compatible software. Even portions of Mac OS X itself, although they do run natively on Intel processors, will likely benefit from further optimization.
Alien Software: Running Classic and Windows
The release of Intel-based Macs is a major milestone for users of Mac OS 9 software.
Intel-based Macs don’t support Mac OS X’s Classic mode. So if you rely on old-school Mac OS 9 applications to get your job done, these new Macs simply aren’t for you. While it’s extremely unlikely that Apple will ever bring back Classic, you may be able to use a Mac emulator, sort of like a Mac equivalent of VirtualPC. At press time, two emulators were already available in “experimental versions” for Intel Macs:
SheepShaver, which emulates a Power Mac, and
Basilisk II, which emulates either a Mac Classic or Mac II. But keep in mind that, since they’re both emulators, they’ll be running those Classic applications at low speed.
Anyone who’s had to keep a spare PC around the office to run Windows programs, on the other hand, might well have rejoiced at the news of Apple’s switch to Intel. But such rejoicing is, for now, premature. There isn’t yet a verified way to install Windows on your Mac and boot into it. (Intel-based Macs use a boot system called EFI, rather than the more traditional BIOS, which makes things harder: Windows XP doesn’t support EFI, although the forthcoming Windows Vista will.) But plenty of clever hackers working hard at the problem, and it’s only a matter of time before someone figures it out.
likely that Microsoft will also update Virtual PC so that it runs on Intel-based Macs. When that happens you can expect that it’ll run Windows at speeds approaching those you’d find on a normal PC. (It won’t be full-speed, however, because Windows will be running inside an application and sharing space with the rest of your Mac.) Other Windows-in-a-box products will probably appear as well. So the future of running Windows stuff on your PC will be bright, eventually.
Different Chips for Different Folks
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.
As a result, it’s difficult to take these first Intel-based Macs and try to extrapolate the entire future of the Mac product line from them. For example, Apple’s forthcoming professional desktop Macs—whatever they’re called, since they almost certainly won’t be called Power Macs—will most likely use a new, high-speed Intel chip designed specifically for desktop PCs. They’ll be fast, and they’ll be designed to trade off a bit in terms of power consumption in exchange for speed. (They’ll also probably support 64-bit processing, which is important to many users of the current Power Mac G5.)
On the other end of things, there’s no way of telling how Apple might take advantage of the forthcoming low-power version of the Core Duo processor, not to mention the Core Duo’s low-power single-core sibling, the Core Solo. All of Intel’s various chips give Apple a large palette to paint with; the big mystery is which colors Steve Jobs will choose to use.
The Last Word
With two iMac models as our only solid evidence of the Mac’s transition to Intel processors, it’s far to soon to issue any final judgments about how the transition will go. There are still lots of huge questions yet to be resolved, including the speed of the MacBook Pro, the speed of Apple’s forthcoming Universal versions of its professional applications, and just how fast the first professional Intel-based desktop Macs will be. (Even more tantalizing are the forthcoming Mac models we can’t even anticipate yet!)
However, these new iMacs do clear up several mysteries and generally make us feel that the Intel transition may be a pretty smooth one. Apple’s claim that the new iMac is 2x faster than the previous model may be debatable, but the fact that it is noticeably faster at almost every native task—and much, much faster at tasks that are multiprocessor-friendly—is indisputable.
We also expect that as time goes on, the apparent speed of these first Intel Macs will increase, thanks to optimizations in Mac OS X and individual applications that take better advantage of multiple processors and specific traits of the Intel-based Mac architecture. And of course, as programs that currently run in Rosetta are updated to be Universal, users will see notable speed improvements.
Most importantly—and when we get deep down into talking about chips and test results, it’s easy to fail to see the forest for the trees—these new systems are Macs, period. That Intel chip embedded deep inside may have a lot of meaning when it comes to the appearance, features, and performance of future Mac models—but you’d never know it from sitting down at one of these iMacs. The Intel revolution has begun, not with a bang, but with a familiar and comforting Mac boot-up chime.