Apple yesterday announced new iMacs and “MacBook” computers using Intel’s power-efficient Intel® Core™ processors (code name “Yonah”). These dual-core chips essentially put two processing units on one chip, as do the current PowerMac G5 models (which don’t yet have Intel-based replacements). This arrangement allegedly gives tremendous computing power at little additional cost, and lower power consumption.
The promise of a 4X faster laptop (according to Apple) is certainly a tremendous advancement for those who need to work with images in the field. However, the sluggish performance of the old PowerPC-based G4 laptops means that this 4X increase will just bring things up to the performance of a mid-range PowerMac G5. Or will it?
Apple’s speed claims are based on industry standard “SPEC” benchmarks that have only modest correlation to real-world performance with real applications, especially with the snail-like hard drives found in today’s laptops (not an Apple-specific issue). The fact that Apple is now quoting SPEC numbers is rather odd; performance figures in the past were quoted for real programs, like Photoshop, or Final Cut Pro (see the PowerMac G5 performance claims). The SPEC numbers are almost certainly more flattering than real-world numbers, and thus make for far better marketing hype.
Furthermore, it’s not clear whether half of the 4X factor is due to the dual-core chip (eg each core 2X as fast and with 2 cores = 4X). If so, then the maximum improvement for many applications will be just 2X, not 4X, since very few applications can actually make full use of two threads (cores), and even those that can do so rarely use both cores fully for more than a brief time.
Even applications such as Adobe Photoshop only briefly fully utilize both CPUs on the PowerMac dual 2.5 GHz desktop—and that’s with 7GB RAM and a 4-way SATA striped RAID where disk I/O speed is sustained at over 220MB/sec. Most of the time CPU usage does not exceed 125% (200% being both CPUs at full use). As shown in the Nikon Capture article, average CPU use by Nikon Capture when batch-processing raw files is somewhere around 125%.
For non-threaded applications (those that will use just one core), a real-world 1.5X performance improvement is likely, with a reasonable possibility of 2X improvement. Much higher bandwidth memory and larger on-chip caches in the new MacBooks should increase the odds for 2X. I hope to be proven wrong, and see the full 4X improvement—but don’t buy a MacBook based on 4X hype just yet.
The foregoing all comes with a huge “gotcha”: Until you update your software to a “native” or “fat” version which is compiled for the Intel chip, it will run in emulation (eg relatively slowly). This is the reality today for Photoshop, all existing raw converters, and virtually all non-Apple software. Except to spend some time, trouble and money to get versions which run natively on the Intel-based machines. Some software vendors will see this as a terrific money-making opportunity, and thus will be on it like flies on a fresh carcass. They will trumpet the 3 or 4 pitiful features they’ve added in order to make you pay for the native version. Return on investment is not greed, and is perfectly reasonable, but keep an eye on this phenomenon and observe—you can learn a lot just by looking.
Finally, all is not good news. Firewire 800 has been eliminated in the new MacBooks. Those using Firewire 800 drives externally will have to obtain a Firewire 800-to-400 cable, and accept slower transfer speeds.
In the long run, this transition offers huge promise, and Apple should be applauded for undertaking it, a transition which few companies are skillful enough to manage, but one which Apple will handle with agility and quality. I look forward to future desktop Intel Macs. A dual dual-core 3.8 GHz box will do just fine.