Mac Musings

The Best Reason for Mac OS X on Intel

Daniel Knight - 2003.01.03

Apple made a wise choice when it picked the Motorola 68000 for the original Macintosh instead of using the same brain damaged Intel processors adopted by IBM and the clones. In truth, IBM had considered the 68000 for the original IBM PC, but they went with the 8088 because it was cheaper, allowed them to use an 8-bit memory path, and simplified migration from CP/M and the Z-80.

PC users have been paying for that decision ever since with an assortment of memory models, processor modes, and operating systems designed for the idiosyncrasies of the Intel architecture.

Apple made another wise choice when they teamed up with IBM and Motorola to create the PowerPC. Although the first generation of 60-80 MHz Power Macs wasn't that much more efficient than the 33-40 MHz Quadras they replaced, they helped Apple make the transition from 680x0 processor, NuBus expansion, costly SCSI drives, and proprietary serial and input ports to the PowerPC family, PCI slots, IDE and ATAPI drives, and industry standard FireWire and USB ports.

And now Apple is polishing their OS X operating system, which is rooted in the standards of BSD Unix.

Processor Efficiency

Not all processors are created equal. The 6502 found in the Apple II line, the Commodore 64, and many other 8-bit PCs held its own against the 4.77 MHz IBM PC in terms of performance. Where they lost out was memory management. With some tricks, the 6510 could juggle more than one 64 KB bank of memory, but the 8088 was designed from the ground up to manage 16 such banks.

Processors got faster, but not a whole lot more efficient. It wasn't until 1990-91 that the fifth generation* 680x0 and x86 processors offered 2-3 times the performance of their predecessors at the same clock speed. This improvement was due to a much larger level 1 cache on the processor and the integration of the FPU (Floating Point Unit) into the CPU.

* I am including the rarely used 80186 and 68010 in this count, even though the '186 was only used in a few PCs and the '010 was never used in any Macintosh. They were part of the genealogy of these chip families.

On the PC side, a 20 MHz 486 was in every way the equal of a 40 MHz 386. But it was even better on the Mac side: The lowly Centris 610 with a 20 MHz '040 outperformed the "wicked fast" Mac IIfx with its 40 MHz 68030 and 68882 FPU - despite the fact that the 68LC040 in the 610 was the "crippled" version that didn't include the FPU.

G3 Processor Efficiency

To my knowledge, we haven't seen such an improvement in processor performance per MHz since then. The 68050 died on the vine, and Apple chose the PowerPC over the 68060. On the PC side, Intel began their Pentium family, which it today's leader in sheer clock speed.

Apple, IBM, and Motorola chose to create more efficient processors, not ones that simply ran faster. Their big breakthrough was the G3, which offered about 50% more power per MHz than the 604 processors it replaced. Apple replaced 300 and 350 MHz Power Macs with 233 and 266 MHz G3 models that offered the same level of performance without the huge power demands and heat buildup of the earlier models.

The G3 went head-to-head with Intel's then-new Pentium II, and the 233 MHz Power Mac G3 easily matched the performance of 400 MHz PII models in most tests. Since then, IBM and Motorola have made the PowerPC still more efficient, while Intel made "negligible horsepower improvements" in the Pentium III and actually created a less efficient design (in terms of computing power per MHz) with the Pentium 4.

Yes, less efficient. With the Pentium 4, Intel has thrown its entire weight behind clock speed at the expense of efficiency - the exact opposite of the way the entire industry has worked from the beginning. Early tests showed that the Pentium 4 had to run one-third faster than a PIII to match its performance, and things haven't changed since then, except for Intel pushing the P4 to 3 GHz.

The Math

The G3 offered 100% more computing power per MHz than the Pentium II. The PIII was a bit more efficient, but the G3 has been improved, and the G4 has some further enhancements. MHz for MHz, today's G3 and G4s probably offer twice power per MHz compared to the old PIII.

Now throw the less efficient Pentium 4 into the equation. If a 1.4 GHz PIII roughly equates to a 700 MHz G3 (just the CPU - we're not looking at the whole computer yet), and if the P4 is one-third less efficient than the PIII, it would take a 1.8 GHz Pentium 4 processor to match the raw horsepower of today's cheapest iBook or eMac.

Using the same equation, the bleeding edge 3 GHz Pentium 4 processor should offer similar power to a 1.1-1.2 GHz PowerPC G4.

Think about it.

Then remember that the Power Mac G4 has two CPUs.

The Real World

There are two problems with this scenario. The first is that we're only looking at the CPU itself. In the real world, the CPU sits on a motherboard, accesses memory over a bus, works with a graphics card and hard drive, and runs an operating system and applications.

The second problem is that no operating system is optimized to run on both Macintosh and Intel hardware. Windows for PowerPC was abandoned long ago, the various Linux distributions offer varying levels of optimization, and Mac OS X has never been released for Intel hardware.

Comparing a Mac running Mac OS X and Microsoft Office v.X with a Pentium 4 PC running Windows XP and Office XP isn't fair. Everyone knows that Microsoft cheats and has used undocumented calls and other shortcuts to give their own applications the advantage over competing software on the Windows platform. They probably don't invest the same level of expertise in making Office for the Mac as efficient as possible.

The best way to compare the performance of Apple's computers with the Dells, Gateways, HPs, and no name clones would be for Apple to create a fully tweaked developer release of Mac OS X for Intel. This need never be a full-fledged consumer product, but it would provide a foundation for comparing the performance of Apple hardware with that of the Wintel world.

Apple could work with various software companies to insure that Pentium releases of their applications were available to run on the developer release. We could finally have Photoshop, iTunes, browser, and other application benchmarks that wouldn't be thrown off by the optimizations of Microsoft Office on Microsoft Windows. It would level the benchmark playing field.

I suspect that we'd find the math used above doesn't apply to the real world. Maybe a 1.25 GHz G4 computer isn't the equivalent of a 3 GHz Pentium 4 machine. Maybe it's more efficient. Maybe it's not. Maybe Apple already knows - and that's why there's no OS X for Intel.

A Mac OS X on Intel developer release would let us know the answers without threatening Microsoft's dominance, since it would not be a general release operating system. Then again, if Apple does eventually decide to move to Intel or AMD processors, a developer release would help pave the way and also help Mac users understand the pros and cons of the different hardware platform.

I don't anticipate or advocate Apple switching from PowerPC to Intel and AMD, but what begins as a way to better compare hardware platforms could eventually grow into a fully developed alternative to Microsoft Windows.