Interview with AMD: Model Numbers, DDR2 and More
Computex is the show to beat all shows when it comes to computer hardware. This year isn't as revolutionary as year's past as there is no tremendous change in the field of PC hardware that hasn't been coming for quite some time already. Dual core processors seem to be top draw, as is the newly announced ATI CrossFire technology, which we previewed ealier today.
I did get a chance though to sit down with a handful of international media to talk with three AMD executives: Vanoy Wang, Henri Richard and K.J. Chou. It was a very informal Q&A session that didn't start with any press notes or any announcements, just AMD and the journalists. The information that was discussed on AMD's marketing and strategies turned out to be quite interesting and may answer a few of those lingering questions in your mind about their product offerings.
AMD and Intel Marketing Strategies
The first topic that I brought up was in regards to AMD's lack of marketing and advertising, particularly in the US, against what seems to be a constant barrage of Intel TV and Internet ads. With a solid performance lead, feature lead and price lead, it would seem that AMD has all the right points to make an effective ad campaign that could address the masses and not just the PC enthusiast market. AMD's response was that the performance and feature advantages that AMD has are already known to those users that are interested or understand those points. The general end user then, may not benefit from such a large scale advertising campaign and thus AMD hasn't delved into it yet.
I next followed with a question regarding AMD's continuing usage of a model number rating system with the launch of their X2 dual core products. It would seem to me that the practice of comparing their processors to a P4 lineup or even a single core Thunderbird part is completely meaningless now, especially with Intel no longer using frequency in their marketing. AMD's response was that after talking with their customers, both the OEMs and even end users, they all liked the system as it was. It was easy to understand and it got the point across: a 5000+ is better than a 4500+. Richard went on to point out that Intel's current situation with their models numbers was forced upon them by the P4 core running out of steam. He again emphasized AMD's initial plans of having an industry wide performance standard (something AMD discussed when the model rating system was initially released) much like the auto industry has horsepower and torque.
The topic of price points on competing dual core products from AMD and Intel came up in the discussion as well. It is known that AMD's lowest priced X2 processor is going to be around $500 at release, almost twice as much as the cheapest Intel Pentium D 820 processor. AMD defended its stance by stating that AMD would eventually have lower priced Athlon X2 processors via the waterfall effect in the future, and by pointing out a fundamental problem with Intel's strategy: price overlap. The price overlap means that Intel has both dual and single core processors at similar price points, so how can Intel effectively market dual core as the "latest and best" when it has higher priced single core products all over the product line as well?
I next asked about something that has come up before: the redesign of the Opteron 1xx series of processors to run on the Socket 939 platform instead of the 940-pin package like the other Opteron processors. To me is seems counter-intuitive to have two processors identical in features and performance to be labeled as different parts (Opteron vs Athlon) and charge different amounts for them. Richard's claim was that they really wanted to bring the lower platform cost of the S939 package to the Opteron and that the additional warranty, burn-in and support that goes along with the Opteron name will still keep the workstation/corporate buyers on the Opteron bandwagon.
During the discussion some random bits about the upcoming AMD technology changes as well as other product lines were addressed. The first of these was about the 2006 change on AMD's side of going from a DDR to a DDR2 memory controller. I was curious whether this change on AMD's processor line was due to the demands of the user or more to the demands of the memory market. After all, if the memory manufacturers are all moving to DDR2 production, then having a DDR1-based product would eventually hurt your price points. Richard responded with a combination of both. The most demanding of all the markets in terms of performance is the server market and moving into DDR2 over speeds of 667 MHz will show dramatic enough performance differences from DDR1 that AMD forsees their customers asking for the technology, so they will have it ready before then. The mobile market on the other hand is more affected by cost, and so AMD needs to have a DDR2 product ready for when the production levels of DDR2 outpace that of DDR1 and thus current cost scales are reversed. Desktop is somewhere in between the two.
On that same line of thought came a question about supporting FB-DIMMs (fully buffered) in a future processor release. Of course no one would comment on that officially yet, they did indicate that with the Opteron line, they are primarily a server chip company and that they MUST support any new technology in the form of I/O, memory or standards that the server market demands. So if the market were to demand a FB-DIMM Opteron, then AMD would be willing to supply it.
It also slipped that with the upcoming DDR2 products we can expect a new refresh on the chipset side (from third party vendors at least) of the AMD platform as well. Nothing specific was said about any vendor in particular but it also leads one to wonder if maybe this is a requirement? Perhaps the new version of HyperTransport will be integrated into the new CPU as well?
When the topic of having a single core FX processor came up, one of the AMD representatives indicated that some "leading titles" out by the end of the year would see some good performance gains on dual core processors and thus we may see a move from single to dual core FX processors sooner than originally thought. This all depends on how the game developers are utilizing the available hardware, however. Anyone else smell some multi-threaded UE3 maybe?
AMD responded to a question on the competitiveness of the Cell processor technology by wondering aloud if the Cell processor was a competitor to AMD at all? After all, it can not run x86 code to which AMD has dedicated itself with its "x86 Everywhere" claims. They did admit that the processor itself was very ingenious technology and that if/when it does become a competitive product to anything in AMD's line up, it will be addressed. Competition is good for the industry after all, right?
Finally, I thought an interesting note that was made was about there being two design teams at AMD: a server and a mobile product team. This may seem against initial reasoning as so much of AMD's processor sales are based on desktop parts. However, Richard claimed that currently what we have is a situation where a server processor (Opteron) was changed and tweaked slightly to be called a desktop processor (Athlon 64). In the not to distant future we might see the reverse in that a mobile processor would be tweaked and changed enough to be called a desktop processor. Perhaps some indiciation of our changing market place? Food for thought, indeed.
I'd like to thank the Taiwan branch of AMD for inviting me to this meeting. And I hope our readers got something useful out of my summary of it as well. As always, feel free to drop me a line with any commments or suggestions.