[RETURN HOME]

Apples and Oranges:  The Cyrix PR Debacle vs. the AMD Athlon Model #

By Joel Hruska

Date: October 12, 2001

 

Van touched on this idea just a few days ago with an article defending the Athlon Model # system, but I wanted to write a brief piece for those confused about the Cyrix PR system—and how the Athlon Model # is vastly different from it.

Cyrix PR Ratings:  Skewed Benchmarking at Best

 Cyrix PR Ratings were originally derived from a single benchmark (Winstone 97).  Under Winstone 97, the Cyrix 6x86MX performed like a Pentium of a higher clockspeed.  Because IBM was looking to market the chip directly against Intel’s fastest offering, their solution was simple.  Rather than focusing on MHz, they would introduce a PR (Performance Rating) system and rate the chip on that.  A 233PR, for example, was a Cyrix 6x86MX running at either 188 MHz or 200 MHz (due to different bus speeds) that Cyrix claimed was equal to a Pentium at 233 MHz.

So why’d the system fall apart?  Why does the PR idea still leave such a bad taste in the mouths of people?  There are several reasons.

  1. The Cyrix PR was based on a single benchmark
  2. Cyrix chips were MISERABLE gaming chips, with FPU power less than 50% that of the Pentium.
  3. Cyrix chips were plagued with incompatibilities when first released—a great deal of software would not run on them until patched.  In addition, the chips ran at faster bus speeds (typically 75 or 83 MHz) that many older motherboards had trouble handling.
  4. Cyrix’s PR system was confusing, with chips of different bus and MHz speeds being ranked identically.  Both a 150 and a 166 MHz chip were a PR200, while a 188 and 200 MHz chip were a PR233.
  5. Cyrix chips of this generation were buggy, hot, unstable, and crash-prone. 

Given those points, its not too hard to see why the PR system failed so badly.  It was confusing, based only on one benchmark, and the chip’s performance was not consistent.  While it did indeed rival faster Intel Pentium and Pentium II processors in business applications, it was demolished in 3D gaming and modeling by even the K6 (no strong FPU contender itself).

Now, let’s take a look at the AMD Athlon Model Number system and see how many of these points continue to apply.

===================================

The AMD Model # System:  Altogether Different

  1. AMD has based the Athlon Model Number on a variety of benchmarks from Office Productivity Suites to 3D Games and advanced modeling software.  No one specific test was used.
  2. Unlike the earlier Cyrix, the Athlon and AthlonXP are excellent gaming processors.
  3. The AthlonXP remains SocketA compatible and AMD has announced most motherboards should require only a BIOS flash to use the CPU correctly.
  4. Unlike Cyrix, AMD’s Model system does not list CPU’s at different speeds with the same rating #.  There is only one bus speed and one equivalent model number.
  5. Although AMD chips do run hot as do their even more power hungry Intel counterparts, they are no more inherently crash-prone or buggy. 

It’s not hard to see the differences here.  The Cyrix PR system was an attempt by IBM to push the Cyrix as a higher performing solution than it was.  The Athlon Model Number, by contrast, has been called conservative if anything.  Of all the various reviews released on the AthlonXP thus far, not a single one I’ve seen has called the XP rating system questionable with regard to its performance claims.  If anything, the model numbers are tilted slightly in Intel’s favor.

I’ll admit, I’m not a fan of using Model Numbers, but I see them as a necessary evil.  There are a number of tech sites attacking AMD for “perverting” the MHz standard and confusing the average customer, but I see this as incredibly one-sided and naïve.  AMD’s model number system may be more confusing than the old MHz standard, but Intel certainly hasn’t helped the situation.  As we’ve demonstrated, the P4’s tremendous inefficiency per clock-cycle (especially when paired with SDRAM) means that 2 Ghz P4 sitting on the shelf isn’t packing nearly the computing power the customer thinks it is.

It’s biased to slam AMD for instituting a “false” marketing standard on the one hand, yet not equally condem Intel for hijacking the old standard to begin with.  As much as MHz was an incomplete measure of a processor’s performance, in the days before the P4 it was at least a consistent measure.  An Athlon 1 Ghz and P3 1 Ghz could be compared and assumed to be nearly identical in speed and consistent in performance throughout a wide variety of benchmarks.  This is no longer the case.  Thanks to the P4 and Intel’s Marketing Department, we have 2 Ghz CPU’s struggling to keep up with 1.4 Ghz counterparts.  This is at least as confusing to the customer as a Model Number.

Model Numbers aren’t perfect—but AMD’s already announced that they are nothing but a temporary standard while a new, more inclusive performance rating system is developed—this one with the help of other major players in the computer industry.  In the meantime, we have Model Numbers on the one hand and misleading MHz on the other.  Slam AMD for the Model numbers if you want—but don’t forget the reason those model numbers are there in the first place—and despite Intel’s protests, it surely wasn’t for the good of the customers, or else we’d be seeing 2 Ghz worth of performance out of that 2 Ghz processor.

===================================

Pssst!  Our Shopping Page has been updated.

===================================

[RETURN HOME]