Chip maker Advanced Micro Devices Inc. continued the momentum generated by Opteron and Athlon 64 into 2005, particularly by beating rival Intel Corp. in releasing dual-core processors.
Company officials expect to continue that push that technological advantage into 2006. AMDs president and COO of microprocessing, Dirk Meyer, recently spoke with eWEEK News Editor Dennis Fisher and Senior Editor Jeffrey Burt in the companys Austin, Texas offices about the commercial client space, meeting customer expectations and Dell Inc.
Can you give us your thoughts as AMD enters 2006?
I think 2005 was a year that, as a company, we were very happy with for the processor business. We came into the year with a lot of expectations for ourselves, particularly around Opteron.
We went into the year feeling that the product was really differentiated versus the competition, and it was a matter of getting the design wins that we felt we had in the bag into the market and ramping with our key customers, and that happened and those products are being pretty well-received in the marketplace.
I like to talk about Opteron as the thin edge of the wedge relative to our penetration into the commercial space, and when we talk about 2006, one of the things we think about is the opportunities were going to have in the client side of the commercial marketplace.
Thats not to say that we dont expect to continue to gain share in the server space: we do. But contrast 06 with 05 in that well be thinking much more about making our commercial client business really start to ramp.
How are you going to do that?
The first thing youve got to do is get design wins, and as we talked about in our analyst conference [Nov. 15] ... about the incremental number of client platforms well have out on the market next year.
The next step is to start to generate familiarity of the platforms across ... the ecosystem, which consists of both end users and the channel from which they buy. The latter is a set of activities weve had in flight ... for the last several quarters, which is developing relationships with systems resellers and VARs, systems integrators, and so on to make better known who we are, what the value proposition for the platforms is, the advantages that the channel players have. … Thats where well continue to focus in the next year.
AMD has spoken often about how it feels it has a technological advantage over Intel right now, and that you expect that to continue through 2006. Where does that confidence come from?
Ill start by talking about discussions weve had with people in the industry about our server products. The server [business] is really our flagship product line for both us and the competition. We started thinking carefully about changing the underlying platform architecture for systems back before we even introduced Opteron, which was introduced in 2003.
Way back in 1999, we observed that … as the processor industry players have been pushing on the microprocessor analog to Moores Law … that came at the cost of increasing power dissipation with each technology generation to the point where power levels are just getting kind of silly—100 watt, 150 watt, just crazy.
The reason for that is that we as an industry were just focused on optimizing and driving up single-threaded performance in the microprocessor, and we said, You know what? Thats not the right bandwagon to stay on. Lets go innovate in three different areas.
One was, lets go change the system architecture to get rid of this shared-bus bottleneck which is kind of terrible for processor systems. Lets add 64 bits to the x86 architecture and lets architect, in the chip and platform right upfront, multicore capabilities.
The competition clearly has now followed us in the middle of those two, which is the 64-bit transition, but hasnt really yet fixed the platform constraints that they have, and those are way more important in terms of unlocking system performance scaling than what technology node youre in at a given point in time.
As we understand what the competition is up to, while theyve got some improvements in the product line, they have yet to overcome what is the fundamental roadblock to good multiprocessor scaling, which are those two things I listed.
Many industry analysts dont believe it will take Intel that long to catch up. What youre saying is that its going to take them at least another year to do so.
Thats why I always go back and say that we started this in 1999. So the products we have now, we can see that the roadmap and the platform that underpins it all [we started] way back in 1999, and thats not just something that you can flip on a dime. As we understand what the other guy is up to, the 06 roadmap doesnt reflect this shift, nor does the 07 roadmap.