Like us, you may be considering the purchase of an Apple MacBook Pro as a way to bring deadline-friendly processing speed to a Mac-based field workflow. For several years, certain models of Apple's Powerbook line have represented a superb blend of features, screen quality and portability; in fact, the overall design of the company's mobile machines has far eclipsed computers we've used from mainstream PC vendors in all key areas, except one: speed. A PowerPC G4-equipped Mac laptop simply can't keep up to a Windows laptop powered by an Intel or AMD processor.
For intensive tasks such as RAW conversion, applying beefy filters like Smart Sharpen in Photoshop CS2 and previewing folders overflowing with 8+ megapixel photos, a Powerbook doesn't keep pace. For example, processing a 16-bit EOS-1Ds Mark II file with Noise Ninja takes about 44 seconds on a Powerbook with a 1.67GHz G4 processor. The same operation on the same file zips by in under 15 seconds on a Toshiba A70/A75 armed with a 3.33GHz Pentium 4. It's this sort of real-world performance difference that has resulted in site co-editor Mike Sturk relying almost entirely upon a Pentium 4-equipped Dell laptop for on-site work, despite the fact he is at heart a Mac guy.
As we noted in an article introducing the MacBook Pro last month, we hope that the switch to Intel processors by Apple will close the speed gap between Mac laptops and laptops from everyone else. But it's really too soon to address whether an Intel-equipped Mac will perform comparably to an Intel (or AMD)-equipped PC, since so few Mac pro imaging applications have yet been turned into versions optimized for the Intel architecture of the latest Macs. For months to come, the majority of Mac applications we rely on to get photo work done will be PowerPC versions, and will operate on an Intel Mac only through the assistance of the Rosetta emulation technology built into the Mac OS.
It's also really too soon to benchmark a MacBook Pro specifically, since they aren't yet shipping and we don't have one. What we do have on hand is a close cousin to the MacBook Pro: an early 2006 iMac with a 2GHz Intel Core Duo processor. As fans of Stephen Colbert's Colbert Report, we applied his standard of truthiness in coming up with the headline for this article. We really wanted to benchmark a MacBook Pro, but couldn't, so we got hold of an iMac instead and having been calling it a MacBook Pro.
There's a method to our madness: prior to Apple's new laptop hitting the streets, we wanted to get a sense of whether - when running Universal Binary versions of an application - the MacBook Pro is going to deliver anything close to the promised 4.5x+ speed increase (using benchmarking software) relative to a Powerbook G4. And whether PowerPC applications pumped through the seamless but speed-robbing Rosetta will perform at least as well as they would on a Powerbook G4. The Intel Core Duo version of the iMac makes for good stand-in, since it contains similar components, including the all-important Intel Core Duo processor and X1600-series graphics card from ATI. So, without further ado, here's what we tested, and the results:
- 15-inch Powerbook G4/1.67GHz with 2GB RAM, ATI Mobility Radeon 9700 graphics with 64MB memory, 1280x854 pixel display, 80GB/5400 RPM hard drive and OS X 10.4.4. This a mid-2005 model.
- 20-inch iMac with Intel Core Duo 2GHz, 2GB RAM, ATI Radeon X1600 graphics with 128MB memory, 1680x1050 pixel display, 250GB/7200 RPM hard drive and OS X 10.4.4. This is an early 2006 model.
As of this writing, the MacBook Pro's Intel Core Duo processor tops out at 1.83GHz, so our 2GHz iMac will almost certainly be a few percentage points faster than a MacBook Pro in tests that exercise the CPU. The hard drive in the MacBook Pro, however, is not at all the same as that found in the iMac. To neutralize that as a performance variable, we used an external 100GB/5400 RPM laptop drive in a FireWire 400 enclosure as the file source or destination, as well as constructed the tests so that there was a minimum of big-file reading and writing. All of that said, we spot-checked several of the tests using the iMac's internal drive as the file destination and saw almost no difference in the results.
In Universal Binary applications, functions that are written for multiple processors show the biggest speed gains. For example, Photo Mechanic's Preview mode utilizes both cores in the Intel Core Duo, which leads to a speed increase approaching 3X, relative to the Powerbook G4 tested. Previewing high-resolution pictures in Photo Mechanic 4.4.1 on the Intel iMac is really zippy. In fact, it feels faster than the 3X bump suggests. Exporting a newly-created JPEG from Photo Mechanic utilizes only one of the Intel processor's cores, and yet we still measured about a 2X speed increase over the G4. It looks like each half of the Core Duo is considerably more powerful than a G4, at least for the sorts of functions that are at the heart of what a program like Photo Mechanic does. We didn't encounter - in Photo Mechanic or elsewhere - a 4.5x speed increase, but then it wasn't all that likely that real world measurements would match Apple's synthetic SPEC benchmarks. Encouragingly, however, the Photo Mechanic results do fall within the 1.7x to 4.1x speed bump range Apple touts in their own application performance testing.
If Apple wants to show the Intel Core Duo processor in its best light right now, and the audience is pro shooters, they might want to load a few copies of Photo Mechanic on their demo MacBook Pros. The Intel-native iPhoto 6 feels faster on the iMac than on the Powerbook G4 in basic tasks like scrolling, switching views and navigating around the program's interface. But as the test results show, the performance jump in most instances is significant but not outstanding. The experience of using Photo Mechanic, by comparison, is really transformed by the Intel Core Duo. The more modest iPhoto gains overall may be explained in part by the fact that some of the tests performed are more dependent on the graphics card than on the CPU to get the processing job done.
In PowerPC applications, functions that thread through both cores in the iMac's Intel Core Duo processor are completed about as fast, or in some cases faster, than by the Powerbook G4. That's an impressive feat, given how much translation Rosetta must have to do to make PowerPC code palatable to an Intel processor. For an Intel-based Mac to fly, it must be running Intel native code. But until then, Rosetta emulation on average allows the performance of a single PowerPC G4 processor. And for certain tasks, such as reducing noise with the multi-processor savvy Noise Ninja, Intel Core Duo processing times are quite a bit shorter than those of the G4.
As we've seen in iMac benchmarking elsewhere, something's up with QuickTime export on Intel Core Duo Macs. Over 15 minutes to export the test slideshow as an MPEG-4, 30 fps movie on the Powerbook is already pretty pokey, but over 25 minutes on the Intel iMac for the same export suggests a performance-hobbling bug is lurking in the QuickTime code. We sure hope so, anyway.
Why is Lego Star Wars among the applications tested? A certain 6-year-old employee of Little Guy Media here assesses the usefulness of a computer by whether it will play this game. Fortunately, he rated the performance of the PowerPC application acceptable on the Intel Core Duo iMac, as long as Shadow Mapping is disabled in the game's preferences. Apparently it's okay to not have Qui-Gon Jinn cast a shadow as he light sabres his way through a swarm of battle droids.
Lego Star Wars is one of many PowerPC applications that just works through the magic of Rosetta, though with a performance hit. In addition to the applications listed, we've also installed and been using Startly QuicKeys 3.1, Adobe GoLive CS2, GretagMacbeth Eye-One Match 3.4, Netopia Timbuktu Pro 8.5, applications in Microsoft Office 2004, Roxio Toast 7.0.2 and StuffIt 10. So far, the list of Rosetta-related hiccups has been short:
- Nikon Capture 4.0 - through 4.4.1 wouldn't install on the iMac. The problem was in the installer, not in the application itself. When we dragged the components that make up Capture 4.4.1 from the various locations on another Mac to those same locations on the iMac, the program itself worked fine.
- Eye-One Match 3.4. The program slows to nearly a halt at several points in the monitor calibration module (not just when advancing past the monitor-type selection screen, as it can on any computer), but in each case the software rights itself and continues on.
Does the Intel Core Duo processor best a Quad G5? Not even close. We performed several of the Photo Mechanic and iPhoto tests on a Power Mac Quad G5/2.5GHz, and the G5-based computer performed a lot faster. For example, previewing 100 EOS-1D Mark II photos in Photo Mechanic is accomplished in about 18 seconds on the Quad G5, compared to 42 seconds for the Intel Core Duo iMac. Importing 20 EOS 5D CR2's into iPhoto takes 40 seconds on the G5, compared to 76 seconds on the Intel.
And this is when the applications are running natively on each processor type. Running PowerPC-coded Photoshop CS2 on both machines widens the gap that much more because, as we've noted, of the required on-the-fly Rosetta translation. For example, applying Smart Sharpen to a 16-bit EOS-1Ds Mark II file takes 33 seconds on the Intel, but only 7.5 seconds on the Quad G5. Noise Ninja filtering of the same file clocks in at 30 seconds on the Intel, and 7.1 seconds on the Quad G5.
The introduction of the Intel Core Duo processor into the Apple lineup means a lot more available horsepower for portable Macs and, to a lesser extent, consumer-geared machines like the iMac (which were already running G5 processors before Apple's Intel revolution began). But this particular Intel processor is not going to unseat four G5 cores running at 2.5GHz for those who need maximum photo processing power on the Mac platform. Especially - but not only - when the application isn't Intel-optimized.
The 20-inch iMac with an Intel Core Duo processor seems to be a fine machine for the digital SLR photographerneeding to carefully balance cost with performance, and who can live with G4-like speed in applications that haven't yet been given the Universal Binary treatment. The screen calibrates well and appears to be on par with our Apple 20 inch Cinema Display in quality, the computer has a sprightly feel when running Intel-native applications, it's almost completely silent in operation, has a good complement of USB 2.0 (3) and FireWire 400 (2) ports, plus built-in Airport Extreme Wi-Fi, Bluetooth and even Gigabit Ethernet. The video out port is DVI, supports mirrored and extended desktop modes and drives the aforementioned Cinema Display for a totally usable two-display setup. The built-in iSight video camera and tiny infrared remote for the somewhat-limited but still useful Front Row software round out an impressive package.
We got this machine on loan as a MacBook Pro simulator of sorts. In using it, however, we've come to see that Apple has stuffed a lot of pro photography goodness into the new Intel Core Duo iMac for shooters considering a desktop Mac purchase but without the budget for a Power Mac.
You can draw your own conclusions from the benchmarks here as to whether a MacBook Pro is likely to deliver enough of a performance improvement over a Powerbook G4 to make a machine switch pay off in the short term. Our take is that with an Intel-optimized application like Photo Mechanic, the MacBook Pro is going to be a fast portable computer, but when running PowerPC pro imaging applications - which will be a necessity for months to come - its processing speed isn't going to be all that different than a Powerbook G4 overall. Until there are a few more key Universal Binary applications, and/or it's demonstrated that an Intel Mac can run a modern flavour of Windows, we're inclined to stay on the MacBook Pro sidelines, using our existing Windows laptops for speed and Powerbook G4's for everything else.