It seems that Apple may have an unlikely, and unintentional ally, in trying to point out the “megahertz myth” to the general public: AMD.
The megahertz myth is the belief that processor speed is the be-all and end-all of a computer’s performance. This, of course, isn’t true. See our “Myth of MHz” series Part 1, 2 and 3 for details.
With the formal launch of its high-end Athlon XP microprocessor series, AMD this week announced an effort to change the way PC processors are gauged. Instead of measuring computer central processing units by the megahertz of clock speeds, AMD is launching an initiative to develop a new reliable metric to judge CPU performance in standard personal computer applications.
AMD maintains that the MHz-only benchmark puts its Athlon CPU series at a disadvantage in the marketplace next to competing PC processors — in particular Intel Corp.’s Pentium 4 and other microprocessors, according to a Silicon Strategies article. The clock speed of personal computer processors has become a key factor in the price tag of personal computers, making the faster CPU more profitable, the story adds.
AMD’s proposal for setting new standards for measuring PC processor power is called the “True Performance Initiative,” or TPI. AMD said it was attempting to assist PC processor customers (PC system designers) to better understand the total benefits of PC performance, Silicon Strategies reports. The company said TPI would help define a new, more accurate measure of processor performance for standard applications, according to the article.
“For most of the PC’s first 20 years, megahertz was a reliable indicator of PC processor performance because the major players used the same architecture for product design, and clock speed was a good proxy for performance,” W.J. (Jerry) Sanders III, chairman and chief executive officer at AMD, told Silicon Strategies. “This is no longer true.”
What’s more, AMD is adopting a new CPU model numbering system that no longer refers to clock speeds in megahertz. The Athlon XP is being introduced in 1800+, 1700+, 1600+ and 1500+ versions. (The Athlon XP 1800+ is supposed to stack up and beat Intel’s 1.8-GHz Pentium 4 processor in terms of performance.)
While AMD is attempting to rally outside support for new TPI benchmarks, Silicon Strategies said it would certainly face opposition in the industry, especially from Intel. “Historically, efforts to set new processor benchmarks have been extremely difficult, if not impossible, because any measurement potentially favors one CPU architecture or design over another,” the article says.
The clock speed issue has dogged AMD before, but the chipmaker intends to address it head-on this time with a marketing campaign that will stress that numbers don’t tell the whole story, sources familiar with the company’s plans told the news service. The issue is more pressing now because AMD will lag Intel by a full 500MHz in clock speed.
In many published reviews, the slower Pentium 4 chips, such as the 1.5GHz, have tested no better in performance than a 1.1GHz or 1.2GHz Athlon. AMD could also take a page from Apple, whose Macs, of course, use PowerPC chips that lag Intel chips in clock speed but compete with or exceed them in most aspects of real-world performance.
In fact, at July’s Macworld New York, Jon Rubsinstein, Apple’s senior vice president of hardware, came onstage during Apple CEO Steve Jobs’ keynote to discuss the “myth of MHz” and talk about the reasons that, overall, the G4 is the fastest chip around for most tasks.
“MHz doesn’t equal performance,” he said. “It’s just a factor.”
Rubenstein said that true performance is affected by architectural tradeoffs in chip design that include the frequency, number of functional units and cache design, all of which play into the performance of the processor. To illustrate one example of this, he compared the pipeline stages in a G4, Pentium 4, Intel’s upcoming Itanium, and Sun UltraSparc 111. The chips have, respectively, 7, 20, 10 and 14 pipeline stages.
The shorter the pipeline, the faster performance of a chip, Rubenstein said. Longer pipelines have more stages, which means more work before getting results, as well as more inefficiencies in the pipeline.