It's high time AMD revamped its PR ratings

AMD recently did not announce that its Thorton core, the 256K cache version of the Barton core will be called AthlonFX. While samples seen by some INQ moles were not marked as such, rest assured that it is going to happen. It simply makes too much sense from a manufacturing standpoint. It makes almost no sense from a product stratification perspective however.

Until about the end of 2002, AMD had two product lines, the Athlon and the Duron. The Duron was a cut down version of the Athlon, having much less L2 cache, and limited to the slower 200MHz FSB. It sold for a lot less than the Athlon, and allowed AMD to pitch the Athlon against the P4, and the Duron against the Celeron, avoiding the direct comparison between the Celeron and the Athlon that Intel was so hot to promote. It worked well, perhaps to well. The Duron was very close to the Athlon in performance, much closer than the P4 based Pentium/Celeron duo. To balance this, AMD made the Duron's top clock speed be about where the lowest Athlon started. Artificial market stratification at its finest.

With Athlon prices low, at the end of 2002, having a cheaper line didn't make much sense, so the Duron was killed rather than moved to Dresden for production. This left AMD with only the Athlon, now called AthlonXP, to compete with both lines. Intel wasted no time telling everyone who would not run away at the sight of a marketing droid how the Celeron competed against the Athlon, and the P4 had no competition. The spin was so thick that Congressional lobbyists looked on in awe, but anyone with a clue knew better. Sadly, most people don't have a clue, so the strategy worked real well.

So, what is AMD to do? Make a second "value" processor line, and push the AthlonXP name up market. This will allow the upcoming Athlon64 to compete against only the highest end P4s and low end Xeons, while the Opteron goes against the high end Xeons, and the Itanium lines. Good strategy on AMDs part. There is however a speedbump.

This speedbump is not the technical side, that part is well thought out. Unlike the Durons, which were made at a separate plant, on a separate process, from a separate die, all to sell for less than the Athlon, the FX is made from the same everything. The FX is just a Barton core XP with half the cache disabled. This allows AMD to salvage XP dies that have more than the acceptable amount of cache errors, and gives them a home for things that don't make the correct speed bin. Win/win, right? Nope.

The reason is the PR ratings scheme that AMD has used so successfully over the past couple of years, and it is one hell of a speedbump for them. AMD rates their chips by performance, not MHz, and have recently shown that a slower clocked chip, with a larger cache, will be listed as a "higher speed" part according to the PR ratings. So, with this precedent set, how do you make a value line?

If you remain consistent with the ratings, as AMD has recently done, both to its credit and detriment, the FX line ends up overlapping with XP line is a serious way, with no performance difference between them according to AMDs own numbers. Do you go for the $200 2800+ XP, or the $115 2900+ FX? AMD has recently put out polls that indicate most average consumers are confused by technology, acronyms and choice in general. Here is some advice to AMD, don't implement a scheme like this! Chaos will ensue at the local retailer, and one out of every five Best Buy computer salespeople will have their head catch on fire.

So, AMD is left with a really good strategy from a production standpoint, a great way to differentiate themselves from the competition, and a monkey on the backs that is awful hard to shake. How do you fix the situation? Revamp the PR rating system.

People both inside and out of AMD have been crying for such a scheme for a long time now, but nothing has emerged. Sure they tossed the whole thing out for the Opteron line, but that is due to the market it is aimed at, a much more educated one than the consumer market. I think AMD should have a three-tiered PR rating system, one that is revamped on a regular basis, say yearly.

It would work in the following way, a PR rating for Value, Mainstream, and 64-bit/performance. Each would include a separate suite of programs aimed at the market segment that the chips are marketed to, and could be updated yearly. In effect you would have a PRv-2003, PRm-2003, and a PR64-2003, nice easy numbers that Joe Consumer could grasp. The PR suite as it stands is in reference to the Tbird core Athlons, even the geeks who bought the scheme have a hard time swallowing that one, it lines up way way to well with the P4 for it to be mere coincidence. Before you send me 200 e-mails linked to the AMD PR whitepaper, I have read it, and sat through no less than 3 presentations on it, I know, I just don't buy it.

That said, you can pick performance numbers and programs for the PRv that match up with the Celeron, say, MS Works, Adobe Photodeluxe, and a few others that come in the average $1800 worth of FREE software that your $399 Celeron came with. You could also choose to benchmark it under the "lesser" MS OS, either Me or XP Home. PRm would be benchmarked with Office, Photoshop, and other mainstream productivity packages, and a game or two, probably not a bleeding edge game suite, but a solid selection. Think along the lines of what could comfortably play on a GeForce 4800 or a Radeon 9600, and you are in the ballpark. The last line, PR64 would be benchmarked on both Windows64 and Linux to reflect it's higher end status. Productivity suites would be more intensive, with movie creation, CAD, and things that bring the computer to its knees. Games would be the bleeding edge here, reflecting the lunatic fringe that buy the top speed bin of chips just because. Think Doom3, Half Life 2, and the rest of the games coming out near the end of the year. Counterstirke, while still fun, is strictly out. 2D platformers, back to the PRv arena.

This scheme, with software updated each year, would allow AMD to segment the market with similar performance chips, and not step on its own toes. While it's got good at that lately, it is not considered a long term growth strategy by most Wall Street analysts. AMD should give its hard working, and very nice, PR staff a break, and adopt a new, tiered PR system. µ