Faster Chips Aren't Better, AMD Insists
washingtonpost.com
By Mike Musgrove
Washington Post Staff Writer
Friday, November 23, 2001; Page E01
As holiday shoppers shift into high gear, finding the fastest computer has gotten a little more confusing.
It used to be simple. All you had to do was compare three numbers -- the processor speed listed on the box, the price listed on the tag and the balance in your bank account -- and purchase accordingly.
Now, however, chipmaker Advanced Micro Devices is arguing that one of those figures is misleading. While many of us look at a system's processor speed measured in megahertz (or, more recently, gigahertz) first and foremost, Sunnyvale, Calif.-based AMD says that number doesn't tell the full story.
The controversy started with Intel's new Pentium 4 processors, which top out at a faster clock rate than the latest AMD Athlon XP processors. Intel's fastest Pentium 4 runs at 2 GHz, while the Athlon XP maxes out at 1.6 GHz.
But the Pentium 4 uses a new architecture that, analysts say, works less efficiently than that of its predecessor. So even though the Pentium 4 runs at a faster clock rate, it's getting slightly less done per clock tick than an AMD chip with the same clock speed -- or even a Pentium III with the same clock speed.
Confused? Advocates of abandoning the megahertz measurement like to use an automotive analogy to clarify the point.
Judging a processor by its megahertz speed is "like buying a car based on the amount of RPMs the engine turns. There is no 'zero-to-60' mark for processors, and there needs to be," said Pat Moorhead, vice president of consumer advocacy for AMD.
"It's like measuring car performance by looking at rpm, not miles per hour," said Greg Joswiak, senior director of hardware marketing at Apple.
This is a battle that Apple, which uses PowerPC chips made by Motorola and IBM, had to fight years ago. These days, though, the megahertz question "has become less of an issue for us," Joswiak said. Convinced that arguments about megahertz make most users' eyes glaze over, the company now pushes unique Mac capabilities rather than PowerPC clock speeds, which are well behind those of Intel's and AMD's hardware.
"There's no iMovie or iDVD for the Wintel platform," he said, referring to popular Mac-only programs.
But that's not a strategy that AMD, which competes directly with Intel for Windows customers, can adopt. That leaves the company at a disadvantage. "What consumers look at is clock speed," said Bob Thompson, author of "PC Hardware in a Nutshell." "It's easier."
The larger truth, which often gets lost in the clock-speed arms race, is that even the slowest processor available today is more than fast enough for many of the day-to-day chores of computing, such as browsing the Web, fetching e-mail, writing letters and managing a digital-music collection. But at the upper end of the market -- that is, where the profit margins are highest -- consumers take the numbers seriously. And even users with more modest ambitions still tend to seize on processor speed as an overall gauge of a system's worthiness.
Presented with the headache of explaining its case, AMD decided to get rid of the number and instead use a figure that purports to tell users how the processor performs in a series of "real-world" benchmark tests.
Instead of identifying an Athlon chip with a certain number of gigahertz, AMD's Athlon XP processors come with a model number that doubles as a performance rating -- from "1500 plus" to "1900 plus." Athlon's cheaper Duron line of chips, however, is still being marketed by clock speeds.
AMD has even arranged for Athlon processors not to flash their clock speed when the computer boots up. They display the model number instead. If you still want to check what your processor's frequency is, you'll have to look it up in the Windows system information listing.
AMD's new measurement comes from benchmark programs that take into account the performance of 34 "real-world" applications in digital media, office productivity and 3D gaming. AMD calls it a "bridge metric" that the company will use until the industry comes up with a more accurate way of measuring performance.
Analysts tend to approve of AMD's point. "It's a marketing ploy, but it's a good one," said Manoj Nadkarni, founder and president of the processor industry investment Web site ChipInvestor.com. "Megahertz alone is not a complete measure of a processor's performance."
A spokesman for Intel said the company does not comment on its competitors, but agreed that there's more than one way to judge a processor. "There are a lot of different ways to measure the performance of a processor -- megahertz is just one," Robert Manetta said.
Here's the fun part, though, as Dan K. Scovel, semiconductor analyst at Needham & Co., pointed out: Different processors do different tasks better or worse compared with each other. A processor that's better for managing large multimedia files might in some cases fall behind a processor optimized for office work. Before the computer industry could adopt benchmarks, it would have to describe which one benchmark to use -- or list a series of benchmarks for different uses.
"There's significant amount of truth to benchmarking as opposed to clock speed, but then, it's a question of how you want to measure it," Scovel said. |
