AMD Athlon XP - Xperience Performance Beyond the Gigahertz Rating

AMD Athlon XP - Xperience Performance Beyond the Gigahertz Rating

AMD's new Athlon XP processors have been released, and while the chips themselves are excellent performers and have several notable features, they unfortunately bring with them a resurrection of the infamous Cyrix/AMD "performance rating". This is a simulated MHz number that does not indicate the actual speed of the chip, but instead indicates an estimate of the approximate performance compared to equivalent competition. That competition, of course, is clearly focused on the relative MHz of an Intel Pentium 4. If this sounds confusing, that's because it is!

Unfortunately I believe this will confuse customers just as it did in the past, and give a bad feeling to consumers when they find that the true MHz rating of their new AMD chip is much less than the speed rating may lead them to believe.

AMD's problem is this: how do you market a chip that is faster than your rival when its clock speed is actually slower? The new AMD Athlon XP 1800+, for example, runs at 1.53GHz and performs about equal (faster in some tests, slower in others) to a 1.8GHz Pentium 4. The Pentium 4 can still maintain this clock speed advantage, however, because the P4 uses a new and different architecture that utilizes a deeper instruction pipeline with more stages. The Pentium 4 has a 20-stage pipeline, which compares to an 11-stage pipeline in the Athlon and a 10-stage pipeline in the Pentium III/Celeron.

The deeper pipeline effectively breaks instructions down into smaller micro-steps, which allows overall higher clock rates to be achieved using the same silicon technology, but it also means that overall fewer instructions can be executed in a single cycle as compared with the Athlon (or Pentium III). This is because if a branch prediction or speculative execution step fails (which happens fairly frequently inside the processor as it attempts to line up instructions in advance) the entire pipeline has to be flushed and refilled. Thus if you compared an Athlon vs. a Pentium III vs. a Pentium 4 all running at the SAME clock speed, both the Athlon and the Pentium III would beat the Pentium 4 running typical benchmarks, as they would execute more instructions in the same number of cycles.

While this would sound bad for the Pentium 4, it really isn't. In fact, it is all part of the design. The deeper 20-stage pipeline in the P4 architecture allows for significantly higher clock speeds to be achieved using the same silicon die process as conventional chips. As an example, the current Athlon XP and Pentium 4 are made using the same 0.18 micron process (which describes the line width of components etched on the chips). The P4's 20-stage pipeline allows the 0.18 micron die process to result in chips running up to 2.0GHz, while the same process achieves only 1.53GHz in the 11-stage Athlon XP, and only 1.13GHz in the 10-stage Pentium III/Celeron. Even though the Pentium 4 executes fewer instructions in each cycle, the overall higher cycling speeds more than make up for the loss of efficiency. Thus the fastest 1.53GHz Athlon XP compares to and can even beat out a 1.8GHz Pentium 4, but a 1.9GHz or 2.0GHz Pentium 4 would still benchmark much more favorably, and even stay on top overall in many tests.

AMD's performance tests show that the 1.53GHz Athlon runs at the same "performance level" as the 1.8GHz Pentium 4, so it has decided to call the new chip an "Athlon XP 1800+", assigning the 1800+ number as a designation of performance relative to Pentium 4 MHz. This type of marketing, where a chip is assigned a number indicating a relative - rather than a true - speed rating, has been tried before without success. Both Cyrix and AMD tried this before, and all it did was leave a bad taste with customers who felt that they were deceived when they found out the true MHz rating of the chips and systems they had purchased. Only time will tell if this marketing scheme will work this time, or if it will backfire as it did in the past.

And there is another problem, the relative comparisons may change, even on the same existing processors. As software is rewritten to understand the deeper 20-stage pipeline of the Pentium 4 (and that is already happening), fewer internal processor prediction and speculation mistakes will be made, and thus there will be fewer times the pipeline is flushed and refilled. Overall instruction efficiency will rise for the Pentium 4, meaning newer software optimized for the deeper pipeline will actually run faster on the same Pentium 4 processor. This may have the effect of tainting the relative comparisons AMD is making now, possibly rendering them inaccurate in the future.

I liken the performance numbers AMD is using to the "wind chill factor" often used by weather reporters in the wintertime. There is of course the REAL temperature, and then there is the so-called "wind chill factor" which is an estimated rating of how cold it "feels". I prefer to know the actual temperature, thank you! The model numbers AMD is using with the new Athlon XP are like a "speed factor", which is supposed to tell you how fast the processor "feels" compared to a Pentium 4.

The following table shows the new Athlon XP model numbers with the actual CPU clock speeds:

Let's see if AMD's claim that a 1.533GHz Athlon XP can match a 1.8GHz Pentium 4 can hold up to analysis.

Checking the Benchmarks

AMD's published benchmarks for the Athlon XP (available online at <athlonxp.amd.com/benchmarks>) do make a strong case for saying that the processor's design, not the clock speed, is the real deciding factor in which systems are faster. Using popular business, graphics, and gaming benchmarks, the system running the AMD Athlon XP model 1800+ beat a comparably-equipped Intel Pentium 4 1.8GHz system by margins ranging from 6% to 18%.

I always tend to be suspicious of vendor-generated benchmark numbers (although AMD used a combination of industry-standard benchmark tests that were verified by an independent firm). However, third-party tests also bear out AMD's claims. Tom's Hardware website tested the 1800+ against Intel's 2GHz Pentium 4 (not the 1.8GHz suggested as the counterpart) and found the AMD processors beating the much faster Intel processor on most non-game benchmarks, and running about the same or slightly slower on game benchmarks. AnandTech's benchmark tests also pitted the entire line of Athlon XP processors against a lineup of Pentium 4 processors running at speeds from 1.5GHz all the way to 2GHz. AnandTech found that the 1600+ (which actually runs at 1.4GHz) was a close competitor for the 2GHz Pentium 4, while the top-of-the line 1800+ (running at 1.53GHz) beat the 2GHz Pentium 4 by about 10% across the board.

The Design Behind the Numbers

What has AMD done to the new Athlon XP to make it equal or even beat Pentium 4 processors running at higher clock speeds? As mentioned at the beginning of this article, it isn't so much what AMD has done, it is the deeper 20-stage pipeline in the Pentium 4 that makes it execute fewer instructions per cycle, while overall allowing it to run at significantly higher speeds.

AMD uses the term "QuantiSpeed" (a marketing term, not a technical term) to refer to the architecture of the Athlon XP. AMD defines this as including:

• a nine-issue superscalar, fully pipelined microarchitecture providing more pathways for instructions to be sent into the execution sections of the CPU. It includes three floating-point execution units, three integer units, and three address calculation units
• a superscalar, fully pipelined floating-point calculation unit for faster operations per clock cycle; this cures a long-time deficiency of AMD processors versus Intel processors
• a hardware data prefetch which gathers the data needed from system memory and places it in the processor's level 1 cache to save time
• improved translation look-aside buffers (TLBs) for storing data where the processor can access it more quickly without duplication or stalling for lack of fresh information

These design improvements wring more work out of each clock cycle, enabling a "slower" Athlon XP to beat a "faster" Pentium 4 processor in doing actual work (and play).

The XP uses the "Palomino" core, which is also shared by the recently announced Athlon 4 mobile (laptop) and Athlon MP (server) processors. Other features shared with other Athlon versions include:

A new (to AMD) chip packaging feature is the use of a thinner, lighter organic compound similar to the chip packaging used by recent Intel processors. It allows for a more efficient layout of electrical components.

There's no question that AMD's claims about its new Athlon XP processors are true: they do beat less-efficient processors which run "faster" in terms of clock speed. The only problem is that the design of the Pentium 4 allows significantly higher clock speeds to be achieved using the same manufacturing process.

However, don't expect the rest of the industry to embrace this "new" way of rating processors without a fight. It's been tried before - and was an abject failure the first time around. Here's why:

PR - Performance Rating or Public Relations?

If you've read recent editions of my book Upgrading and Repairing PCs, you already understand that clock speed is a long way from being the most important factor that makes one computer faster than another. Back in the mid-1990s, for example, I watched as the industry began to move away from 486-based systems to Pentium-based systems. At the end of the 486 era, clock speeds on 486-derived processors like AMD's 5x86 had reached 133MHz, compared to just 75MHz for a typical entry-level Pentium processor. However, the Pentium 75MHz was actually equal or superior to the "faster" 133MHz processor because the Pentium 75 had a faster front-side-bus speed and wider data bus than the 486-derived 5x86-133MHz processor.

It would seem, then, that the industry should already be primed for a rational discussion of actual CPU performance instead of being hypnotized by CPU clock speeds. Ironically, AMD's own pioneering Pentium rival, the K5, introduced in 1996, helped stop this sensible idea before it could get started. The K5 used an ancestor of AMD's current + ratings, the PR (Performance Rating) system. Under the PR system (also adopted by Cyrix to describe its 6x86 and 6x86MX processors), an AMD K5 CPU identified as a "PR166" ran at an actual clock speed of 116.66MHz; AMD also used this system with slower versions. AMD was claiming then, as its new Athlon model rating system does today, that a slower chip can perform as well as a faster chip. The AMD K5, unfortunately, couldn't live up to its name. It was widely derided as slow and not all that compatible with the Pentium. It failed, but AMD's second try at a Pentium rival, the K6 series, succeeded admirably, and has led to AMD's current line of outstanding processors.

Why did the old "PR" system fail? It wasn't really because the clock speed didn't match the model number. Instead, it was because the chip didn't perform up to the level its model number suggested. That appears not to be the case with the new Athlon XP, however Intel claims that as software is rewritten for the deeper 20-stage pipeline in the Pentium 4, the instruction efficiency will increase further adding to the performance of the Pentium 4. Even so, using current software, AMD's Athlon XP 1800+ has proven itself to be more than equal to handling a Pentium 4 1.8GHz, and even keeps up with a 2GHz version in some tests, while slower models are equally competent at matching or beating slower Pentium 4 models.

How AMD Plans to Make Its Point

Have you noticed how most recent systems display the clock speed of the processor during the initial boot cycle if you have the system configuration option displayed at boot time? Windows XP also displays the CPU clock speed on the General tab of the System Properties sheet. AMD would prefer that systems no longer "tattle" on the processor in this way. In fact, AMD will not recommend or validate motherboards using the Athlon XP if they display the processor's actual clock speed. In the future, the curious will need to use a third-party program such as SiSoft Sandra or Norton Diagnostics to find out the true clock speed of their processors. Unfortunately this will add to the confusion on the part of consumers, who may feel "ripped off" when they find out that their new "1800+" chip is really running at only 1530MHz.

Conclusion

Computer users who've been around for awhile understand the differences between clock speed and real performance, thanks to accurate system benchmarks and critical commentary about the video card, hard drive and memory choices made by vendors which can slow down systems. However, with Intel's Pentium 4 making major inroads into lower-cost systems (albeit by matching the Pentium 4 with slow PC133 SDRAM at a big loss in performance) and major computer manufacturers turning their backs on AMD, it may be much harder for AMD to make its point with the many novice computer buyers out there.

Web Resources

AMD's official web site for the Athlon XP, including technical papers and benchmarks, is available at:
athlonxp.amd.com

Find a motherboard that works with various Athlon models, including the new Athlon XP series, with AMD's motherboard/processor search engine, available at:
www.amd.com/us-en/Processors/TechnicalResources/0,,30
_182_869_1039^2338,00.html

Tom's Hardware provides a detailed look at the Athlon XP 1800+ versus the Pentium 4 at:
www6.tomshardware.com/cpu/01q4/011009/index.html

AnandTech's full-line Athlon XP versus Pentium 4 shootout is at:
www.anandtech.com/cpu/showdoc.html?i=1543

If you're wondering how the Athlon XP 1800+ does against a "plain" 1.4GHz Athlon, check out the AMDZone review at:
www.amdzone.com/articleview.cfm?ArticleID=837

Copyright©2002 Pearson Education. All rights reserved.