[RETURN HOME]

In Defense of AMD's Model Rating System

By Van Smith

Date: October 6, 2001

Intel and its marketing minions have decided to hitch the chip company's fortunes onto the back of its young and controversial flagship, the Pentium 4.  With its deep pipeline, the P4 chip can reach clock speeds that are currently unsurpassed and Intel is opening up its purse strings to let everyone know about it.  What the Santa Clara chip company doesn't want you to know is that the chip is perhaps one of the most anemic offerings, tick-for-tick, of any modern processor and actually and abruptly reverses the natural trends in the evolution of CPU development -- even in its own product line -- towards greater processing efficiency.

We demonstrated this one step forwards for marketing, one giant leap backwards for mankind in a recent article.  In it, we showed what every insider in the industry already knew: that the Pentium 4 is much, much slower at the same clock speed than its predecessor, the Pentium III.

In fact, Intel has had to artificially retard the Pentium III so that the chip does not run roughshod over much faster clocked Pentium 4s.

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

OEMs Deep-Six Intel's Slothful 2 GHz Server P4

The Intel Pentium 4 is so much slower than the Pentium III that OEMs recently forced Intel to kill the expensive 2 GHz P4 Xeon server part since it demonstrably performed worse in dual processor configurations than the 1.2 GHz Pentium III. 

Of course there was a level of collusion on Intel's side since widespread publicity of this fact would dilute the chipmaker's goal of marketing megahertz over MIPS.

But Intel is caught in its own trap.  Its most powerful server chip, the Itanium, runs at 800 MHz or less.  By its own Pentium 4 positioned megahertz measure, the Itanium looks like a woefully slow -- and, at over $3,300 per chip, frightfully expensive -- microprocessor.

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

Intel Plays Dirty

The Santa Clara, California chip colossus is, by simple virtue of its size, a formidable competitor.  But Intel is not satisfied with playing fairly and has been engaged in covert, subversive actions for many years.  Some of those are as follows:

Intel "buys" OEMs.  With its "Intel Inside" campaign, Intel subsidizes great portions of participating companies' advertising expenses.  By dropping all chips from rival AMD, these companies are then given sweetheart discounts to Intel's products.  These tactics have been particularly successful in today's soft and unsteady market and many large OEMs have flocked under Intel's wing for cover.  Not surprisingly, Intel's loyalest of the loyal, Dell Computers, has been able to weather the market drought better than any other OEM, thanks in great part to its "special" relationship with Intel.

Intel "buys" the media.  The chip monster has tentacles that reach deeply into the media.  Intel funded CNET's rise.  CNET, now the proud owner of ZDNN (Ziff-Davis itself has often been called into question for its objectivity) was a startup funded by Intel and is now the largest and most powerful mainstream news and information site focusing on high technology.  CNET, the media organ created in part by Intel, serves as a giant rose colored lens through which the entire world looks through to view Intel.  A little over a year ago I confronted Intel's flustered Andy Grove with the obvious conflict of interest regarding Intel's low-key financial connections with CNET, the organization that the public now depends upon for news and analysis of Intel and its competitors.  Although Dr. Grove admitted no wrong doing, several months after this exchange Intel dumped all of it financial interests in CNET, whose stock has since tanked from over $30 per share to about $3.5 per share.

Intel pressures websites.  Of course there are some sites that are widely considered sell-outs to Intel.  The recently gutted SharkyExtreme is an often cited example.  More recently AnandTech and even my old haunt, Tom's Hardware Guide, have started directly carrying Intel advertisements.  Although this has long been considered a no fly zone for websites since it can obviously compromise a site's image as well as directly influence its objectivity, with today's severe crunch in ad revenue some sites have tapped this "dirty" revenue source.  And Intel's pressure doesn't end here.  Even hardware sites that try to remain above Intel's direct financial involvement are frequently pressured by the company.  For instance, when VHJ first approached Intel for review hardware, Intel effectively told us that because we have a reputation for sometimes being critical of the company, that only if we write what they deem acceptable, will they then consider sending us review products.  In other words, in not so subtle terms, Intel communicated to us that we would only receive review hardware from them if we demonstrated in our articles a consistently friendly tone towards Intel and its products.  No other company puts such stipulations on review hardware and Intel's tactic hardly creates an atmosphere conducive to objective reviewing.

Intel controls benchmarks.  The connection between Intel and the leading benchmark organization, BAPCo, is well known inside the industry and is a topic we have written about in the past.  In a fashion as odious as its media connections, by controlling the tools through which reviewers evaluate the merits of competing microprocessors, Intel is able to ensure that their products are given a leg up on the competition.  Benchmark development is an area susceptible to infinite permutations of corruption and bias.  It is for this reason that we remain a strong proponent of free open source benchmarking.  Only through the widespread availability of a benchmark's source code can true verifiability exist.

Intel is cornering the market on compiler technology.  This gives the chip sauropod great advantages in tests like SPEC which allow participants to compile the benchmark using whatever tools they have available.

Intel bullies other companies.  Threats of lawsuits and/or withholding technology have become synonymous with "Intel."  Even though the company often loses its lawsuits -- for instance it sued Cyrix five times and lost every time -- these lawsuits can be debilitating to those companies it attacks, driving down stock prices, creating market uncertainty, and costing companies big bucks in legal defense.  In many ways, Intel has become a lawyer's best friend.

Intel initiates strategies designed to rope in and destroy temporary allies.  AC'97 is a good example where Intel pulled in a number of established companies only to subvert the entire sound card industry.

Intel utilizes great leverage into the financial analyst community.  For example, this enabled it, in the beginning pangs of its current market debacle, to pressure financial analysts to revise Intel's earnings forecasts upwards by including a gigantic one time sell off of Micron stock.  Analysts that did not comply were threatened with not being carried by First Call/Thompson Financial.  It is no wonder that Intel's stock continues to outperform AMD's, despite Intel's collapsed margins and limited success with the Pentium 4.

Intel lies.  It is hard to put it any other way especially when confronted with statements like this one we mentioned on September 19th:

The 845 chipset provides a significant performance improvement over Intel's 815 chipset, which uses a Pentium III chip, Intel said. In comparative tests on mainstream applications, a 1.8GHz P4 using the 845 chipset ran between three to four times faster than a 1GHz PIII chipset using the 815 chipset, said Jeff Austin, marketing manager for Intel's desktop platform group.

Contrary to this statement, it is easily demonstrated that the i845 mated with SDRAM brings the performance of the 2 GHz Pentium 4 down to the level of a 1 GHz PIII on many tasks.

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

AMD's Challenge

The cause of much of Intel's grief has been a surging AMD.  Even in an extremely unfavorable depressed market thrust even further downwards by the recent terrorist acts, AMD's CPU shipments have managed to stay at record levels.  However, Intel, with its deep pockets, has launched a price war that has driven down AMD's average selling price per MPU and has resulted in a larger than expected third quarter loss.

AMD stands on the brink of launching a revised and enhanced Athlon.  In doing so the company also plans to introduce a performance rating system that has already brought the company much criticism.  Instead of marketing their new XP processors according to clock speed, AMD will label their chips by a "Model Number" that is normalized to the clock speed performance of the older Athlon "Thunderbird" architecture (despite rumors to the contrary, the rating system is not based on the Pentium 4).

Certainly, unlike the Intel Pentium 4, the AMD Athlon XP will bring an advancement in processing efficiency over its predecessor, the Thunderbird Athlon.  The Thunderbird, in itself, is superior on a clock-for-clock basis over the Pentium III and therefore far beyond the Pentium 4.

What this means is that the Athlon XP will enjoy an even greater lead in processing efficiency over the Pentium 4.

For years consumers have used megahertz as a measure of relative merit between processors.  Before the Pentium 4, this was not completely without worth.  Now, thanks to Intel, megahertz mean nothing.  A 1.533 GHz AMD Athlon XP will, performance-wise, destroy an Intel 1.8 GHz Pentium 4 in almost every measurable way.

Since megahertz mean nothing, consumers need a quick and easy number to summarize relative performance characteristics of competing processors.  It would be nice if this means was produced by a dispassionate outside source, but since this is not forthcoming, AMD has introduced its own system.

Certainly some are going to interpret AMD's Model Number nomenclature as megahertz, but is that any worse than Intel's deliberate attempt to try to confuse the Pentium 4's megahertz with the Athlon's megahertz? 

Certainly some are going to confuse -- and already have -- that AMD's Model rating is based on the Pentium 4 instead of the Athlon Thunderbird, but since this error actually favors Intel it is not possible to argue deception on AMD's part.

A simple, concise relative performance rating is needed in the computer industry, and AMD's Model Number nomenclature, though not perfect, is a good start.  With it, consumers will have a better idea of the real power of AMD's chips in relation to Intel's.

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

Pssst!  Our Shopping Page has been updated.

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

[RETURN HOME]