   |
|||
|
|||
|
:
News
: Archives
|
|
AMD Q3-2005 |
|
|
And Why some Analysts May Need to do Their Homework |
|||
|
|
|||
|
by Josh Walrath |
|||
|
This is not to say that the engineers at Intel are going about this in a blind manner. The GTL+ based infrastructure is well supported and well known, and it is hard to change such an underlying architecture in short order. AMD bit the bullet with the Athlon 64 architecture, and really only in the past year have they achieved a level of success that is unprecedented in the company's history. It has been a long road for AMD to achieve this level of support for HyperTransport and their Athlon 64 architecture. Intel will have to do the same when they transition their desktop products to a more scalable architecture. Intel is going to be starting this transition in early 2007 with their server level chips. At that time we will see Intel implement an integrated memory controller into their chips, as well as a very HyperTransport-like scalable link architecture. Intel of course will not join the HyperTransport Consortium and utilize that technology, as it is probably not in Intel's best interest to contribute technology to a consortium that includes its arch-rival. To me, this appears to be a transition to an architecture that closely mimics what AMD did with the Opteron two years ago. Once we take all of this into account, we see that most likely there will be a lot of parity between AMD and Intel with their architectures, and AMD could still very well come on top due to its forward looking base architecture. Intel is a strong competitor though, and I firmly believe that discarding the Netburst architecture is the best move they have made since the introduction of the Pentium !!!. The big question that we should be asking though is when Intel will transition their desktop products to their new architecture involving direct interconnect and integrated memory controllers? I cannot see Intel making that move on the desktop or mobile platform for at least a year and a half after the Conroe products have hit the shelves. That means in Spring of 2008 we can expect to see an Intel desktop part that finally sheds the FSB architecture. Then we need to ask ourselves where exactly will AMD be at that time? They certainly won't be standing still. This is no longer a case of David vs. Goliath, as both companies have a very even footing in terms of design and process technology. Where Intel does rule though is the ability to produce and ship many millions more processors than AMD. AMD is looking to possibly convert Fab 30 to 65 nm if that is technically feasible (which it might very well be), and if that is the case then AMD could possibly supply up to 150 million processors a year, which comes out to be around 40% of the current market. Remember, that is the current market. 3.5 billion people in this world do not have a computer, and that is a huge untapped market for PC's. So, for the time being Intel will continue to hold dominance over the PC market, but that does not mean that AMD will not be able to compete and produce a profit. Far from it in fact. AMD has some very talented and focused engineers that have proven themselves to be innovators in processor design. They will continue with solid products, and I can foresee them having a lot of success in the next three years if they continue as they have. AMD has the possibility of addressing up to 30% of the current market, but we will likely see their overall marketshare grow a percentage point at a time, and I highly doubt they will reach that 30% number in the next year and a half. Still, there is a lot of growth potential with AMD, and if management keeps costs under control while retaining their productive employees, then I can see them continue being a strong competitor to Intel. Update October 13, 2005 I think I had misread one big thing in particular while researching this article, and that may have a negative effect on a few of my arguments. This big thing is that Fab 36 will not be going full bore for some time. It will start limited wafer production on Friday when it is officially opened, but it will be far less than 1000 wafer starts per week. AMD does not expect the full ramp to occur until January, and they are expecting around 2000 wspw after they are fully ramped. Currently Fab 30 can do around 25,000 wafer starts per month if it is pushed to its limit. This could have several consequences. The first being that we probably won't see 65 nm parts from AMD until Spring 2006. There is still a lot of work and validation to be done on Fab 36, and I believe there will be a trickle of 65 nm parts coming off that line, but nothing that we could probably get excited about. After it is ramped in January (and who knows when the ramp up will be truly finished), then we will have a better idea when we will see mass produced parts from that line. Some things have not changed though. We will still see FX-59 be delivered in January, and we could possibly see a new dual core product running at 2.6 GHz. I am still somewhat confused about when exactly AMD will be introducing their new sockets, and therefore their support with DDR-2. I believe that the current memory controller can handle DDR-2, but I will need confirmation on that. If this is the case, then we will see much of the current lineup be transferred to the new sockets in short order. I still expect to see new 65 nm parts by Spring of 2006 though. AMD is not terribly worried about what is coming down the road from Intel. The Conroe parts are going to be significantly better than the Netburst based products that are currently available, but most engineers at AMD who know a few things about the industry are not worried that AMD will be unable to compete with these products. From looking at what AMD has done since 1997, and how they have reacted to Intel (and in the past few years have caused Intel to react to them) I think that AMD will be in good shape once these new products from Intel hit the market. We can expect at least one new core to be introduced between now and late Summer 2006.
If you have found this article interesting or a great help, please donate to this site.
Copyright 1999-2005 PenStar Systems, LLC. |
|||
