MacEdition’s Soup: Click here to return to MacEdition Home Page

Chip Chat! Water, water everywhere but not a drop to drink!

By SoupIsGood Food, (soup@macedition.com), 3 January 2001

“500-MHz G4’s aren’t fast enough!” you cry, “Aren’t there any other chips Apple could use in the Mac?”

Sure there are, Speedy. Microprocessors are everywhere, from dishwashers and cars to supersonic jets and megamillion-dollar gene sequencers. The microprocessor market is very competitive and vibrant, and there are a lot of options out there. But would any of them make the Mac the powermongering monster we all crave?

Keeping it in the family

Apple uses two processors, which they call the G3 and G4. These really are the PowerPC 750 from IBM, and the PowerPC 7400 from Motorola. These two chips are part of a much larger family of chips called “PowerPC”, and there are dozens of models to fill every customer’s needs ... except Apple’s (or so it seems at times). IBM makes big computers. The fastest computer in the world, IBM’s ASCI White, runs on PowerPC chips, and so do their 64-bit mainframe-class systems, the RS/6000 and AS/400. (Now called pSeries and iSeries, respectively.) These big bruisers use RS-64 III chips, designed for massive throughput, moving large amounts of data all day long, and the POWER3 chip, which was designed with raw and unrefined speed in mind. The upcoming POWER4 is going to be the fastest slab of silicon ever slapped into a desktop box, and it’s a PowerPC chip at heart, with more than a little in common with the G3 IBM sells to Apple. Matter of fact, programs that ran on older IBM systems running the PowerPC 604 – just like the one in the PowerMac 9500 – will run with no problems on the POWER4-based system, or any of the 64-bit microprocessors in IBM’s arsenal. The problem is that these are high-end servers and UNIX workstations that can cost more than $100,000, and certainly not less than fifteen grand.

Since Apple needs fast personal computers that cost less than a new family sedan with the sunroof package, the POWER4 isn’t going to help ease the perceived PowerPC performance gap.

Biggest and baddest ain’t the bestest

Where else can we go? Intel and x86 certainly isn’t a good option. Or even a workable one. IBM isn’t the only player in the big-number speed sweepstakes, though. The legendary Alpha comes immediately to mind. This big bruiser of a RISC chip has been at the top of every performance benchmark you care to point out for the past eight years, even ones that traditionally favor Intel systems. On occasion, the Alpha has been temporarily surpassed by one of its rivals at the top end, only to make a huge comeback and drown the competition with its sheer number-crunching capability.

Designed and mass produced originally by Digital Equipment Corporation, and then spun off into its own company, Alpha Processor Inc. now has their speed demons manufactured for them by both Samsung and Intel. The new breed of Alpha, the 21264, and the older-but-still-really-fast 21164 are deployed by a number of companies, including Compaq, in various capacities. This means they design Alphas that are fast, and also Alphas that are, for Unix systems, affordable. With Apple’s economies of scale, it’s easy to envision 1.5-gigahertz fire-breathers shutting down for good the Wintel competition in the performance sweepstakes.

There’s a catch, though. There always is.

The Alpha processor is designed for speed above all else, and it gets speed. The problem is that in the pursuit of speed, the Alpha is designed to suck down prodigious amounts of power, and generate enormous amounts of heat. (70 watts at 750 MHz!) This isn’t an issue in a desktop box, but it’s an absolute dealbreaker for notebooks. Also, “affordable” for Unix is “expensive” for most users. Finding an Alpha processor that will both deliver acceptable speed and be cheap enough for the iMac is a tricky proposition.

The Alpha certainly is fast enough to take the PowerPC’s place, but Apple can’t put them into PowerBooks, iBooks or low-end iMacs due to cost or power issues.

Another competitor in the high performance market is the SPARC family of processors, and it can trace its lineage all the way back to the very beginnings of RISC processor design. These are the chips that run Sun workstations and servers, and they’ve just brought out the next generation UltraSPARC III, a 64-bit chip with high-end computing applications in mind. Unfortunately, it suffers the same problems that the Alpha does. Not only is it pricey, it uses up too much power for reasonable battery life or cooling in notebook computers. A company called Tadpole has a notebook that uses a version of the older chip designed for inexpensive and power-conscious applications, the UltraSPARC IIi. This system tops out at 400 MHz, though, and costs as much as a nice Volkswagen with performance that doesn’t match the current crop of PowerBooks.

Little RISCs don’t pay off big

Are there cheaper chip alternatives? Ones that will fit nicely into notebooks, and that will only take dainty sips from the battery instead of big gulps? Certainly! These are the RISC chips that you find in your astonishingly capable game consoles, in aircraft control systems and in specialized Internet routing hardware. This is usually called the “embedded market”, and the field is broad and deep. The Newton MessagePads used a family of this type of processor, the ARM, which has a pretty good balance between power consumption, low cost, and speed. It also tops out at 266 MHz, so the sacrifices made for speed are a little too steep for most Mac users to bear. The Hitachi SuperH series of processors, like the ones found in the Sega Dreamcast, are another example of the low-buck, high-performance breed that sports a 64-bit chip architecture with cheap and portable devices in mind. Unfortunately it also pales in speed comparisons to the G3 and Pentium III. It’s more than enough for day-to-day Web browsing and word processing, but more demanding applications like DTP, audio engineering and video editing need a skosh more brute force than either ARM or SuperH can provide.

The last chip architecture we can look at is called MIPS, and it can be found both in hand-held PDAs and in Silicon Graphics supercomputers. The big problem is that the high-end chips, the R1x000 series, is designed jointly with SGI. SGI, for incredibly stupid reasons beyond the scope of this article, spun off MIPS into its own company in advance of abandoning the chip family, leaving the MIPS architecture with just the R5x00 series of processors. These are like the Hitachi SH in scope and performance, and can even be found in Sony’s PlayStation. Apple could, if they chose, license the chip design from MIPS and take over where SGI left off, but that would mean a huge investment into chip design R&D, as well as having to find companies to manufacture these chips. On the other hand, companies like SiByte are running with licensed MIPS cores, and working on high-performance solutions ... geared toward networking hardware. Whether or not this type of chip can be effectively rolled into a personal computer remains to be seen.

This is the hell that Apple is in. They need a processor that is fast enough to stay a step ahead of Wintel, efficient enough to put into a PowerBook and cheap enough to put in an indigo iMac. There really aren’t any other options in the chip field that can combine reasonably high performance with low cost and efficient power consumption apart from x86. Unlike even x86, the PowerPC is one of only two chip architectures that can scale from tiny chips designed to be stuffed into household appliances, all the way up to horrifically expensive and powerful chips designed for titanic computing tasks. Apple’s needs fall squarely in the middle of this range.

So, PowerPC is still Apple’s best bet, provided they can convince their AIM partners to put a little more effort into the middle of the price/performance scale, where Apple needs it most.

Is there hope for the AIM alliance? Will we be exiled to Wintel hell? Stay tuned! Our own Don Granberry has his sleeves rolled up and is ready to address the bleeding edge of PowerPC present and future.

E-mail this story to a friend