Into the Valley AMD has enjoyed a magical run over the past two years. Its Opteron server chip moved from being a plaything in IBM's server line to becoming a driver of sales for the likes of HP and Sun Microsystems and a major force at a host of other companies. Start-ups have bet their futures on the chip, as have venerable hardware makers like Cray. A whole crop of third-party suppliers such as Pathscale and DRC have cropped up to support Opteron's Hypertransport technology, giving AMD an edge over Intel.
This server chip bounty, however, will come under siege later this year when Intel releases a new fleet of processors designed with catching up to AMD in mind. No one yet knows exactly how well Intel will compete. It's slideware versus hardware at this point.
That said, AMD head of sales and marketing Henri Richard knows that Intel means business. In this interview, he discusses AMD's plan of attack in the coming months and his strategy for keeping interest in Opteron strong. In addition, Richard discusses how HP, Sun, IBM and Google stack up on the Opteron landscape and makes a pledge that AMD will win Dell as a customer.
Last but not least, Richard tells us why the anti-trust lawsuit against Intel was the right thing to do.
El Reg: Intel has promised customers new server and desktop chips in the second half of this year that will put some serious pressure on you. How quickly do you expect the new Intel gear to come to market?
HR: Their story is not coherent.
On the one hand, they are making a lot of noise about technology they are not shipping. And, on the other hand, they are saying that that technology will be 20 per cent of shipments by the end of the year. That's not exactly a stellar ramp.
What remains to be seen is if they will put a few parts out there to slow down the market, or do they have a much better manufacturing plan, and they want to attack by surprise? Or is the reality that it's a lot of vaporware, and they don't have that many parts?
I don't know. I don't think anybody knows expect for Intel.
I certainly am not betting on them failing. That is not a good strategy. I respect the competition and assume they will deliver on their promise.
El Reg: When this performance gap between Opteron and Xeon does level out, will you be able to attract customers at the same rate as you have in the past?
HR: I think even if we didn't attract new customers, for whatever reason, then going deep and wide into the current customers is plenty to keep us busy for years to come. We are now in 90 per cent of the Fortune 100 and 57 per cent of the Fortune 1000. We have plenty of work to do.
Beyond that, end user decisions are not driven strictly by benchmarks. Depending on the workloads at a given customers, it's possible that they might use Intel technology for one part of their data center and AMD in another area based on where the two architectures shine the best.
We are going to maintain the lead in many areas, and we'll see whether there is a tie or not in other areas. It is a little too early to tell.
If you look at how OEMs are planning their roadmaps as a good indication of relative competitiveness and perception of our ability to win in the marketplace, I think we have a very solid second half. If you look at where we will be at the end of December relative to where we are at the end of June, we will have many more platforms out there.
I have been saying that there is a direct correlation between the number of platforms and out market share.
[Hear Richard talk more about the jockeying for position between AMD and Intel here.]
El Reg: Are you disappointed with IBM's limited use of Opteron?
HR: Absolutely not. IBM is an excellent partner. Each one of our partners have a completely different strategy with Opteron. It's good.
El Reg: What do you think IBM's hang up is with not broadening its Opteron lineup more aggressively?
HR: IBM had to make a reasonable decision in terms of amortizing past investments.
El Reg: Their X3 chipset for Xeons?
So, of course, we would like to see more Opteron presence in the IBM lineup. But when you have spent a significant amount of money in developing proprietary technology, it's logical for you to try to get that investment back.
IBM has a different strategy, but they are an excellent partner, and I am very confident about the future of Opteron inside the IBM corporation.
El Reg: Our major issue with AMD at this point is your reluctance to talk about what is coming down the road. You were more than happy to talk about 64-bit extensions early and dual-core parts early. Why aren't you being more specific about what you have to do to counter Intel's improvements?
HR: First off, because we don't feel as much as the outside world does a need to counter. Right now it is more asking Intel what they are going to do to counter AMD than the opposite.
We are in front and gaining share.
El Reg: Why admit then that they will put more pressure on you and talk about how Intel will catch up in some areas. Why not talk about what you have cooking instead?
HR: I am not in favor of hyping the future too much. We have a lot of good parts to sell today.
I think Intel hurt themselves in the first quarter pretty badly. What they essentially said to any faithful Intel customer is, "wait". If you are a smart Intel customer, why in the world would you buy anything in Q1, Q2 and even Q3. Why? You better wait.
They told the entire world that their current products are crap, and the next thing is much better.
I don't have any intention to go tell my customers that the stuff we have now is not as good as what we'll have tomorrow.
To your point though, I think we have a been a little quiet, and our intention is to make a lot of noise at our June analyst conference.
[Hear Richard dive deeper into what he sees as Intel's woes here.]
El Reg: Rank these companies in order of who is the largest Opteron customer. Sun, HP, IBM and Google.
HR: No comment.
El Reg: Can you tell us who is the largest Opteron customer?
HR: No, I don't think we can do that. I'll give you a hint. You can look at those various partners' product lineups and that gives you a sense.
El Reg: We think HP is the largest customer.
HR: That's probably a realistic expectation.
El Reg: How big a customer is Google?
HR: Well, you know, we have never made any comments about Google being a customer. If they were our customer, they obviously would be very big.
El Reg: Are you denying that they're a customer?
HR: Well, we don't make comments on customers.
El Reg: You guys put out customer win press releases all the time.
HR: We have a lot of embedded customers around the world that don't want us to disclose the nature of their relationship. I am not saying that is the case with Google.
El Reg: There seems to be this notion brewing that x86 SMPs could become a bigger part of the overall server market. What's your take on that idea?
HR: We believe that it's a bit of a chicken and egg situation.
The reason that 4P and above isn't a bigger part of the action right now is because Intel was alone and their architecture wasn't scaling. The pricing structure was not helping customers go there either.
Until we came into the market, it was cheaper to do two 2Ps than one 4P. In some cases, it was better performance.
I think AMD and Opteron have changed that.
By nature in the enterprise, some of those changes don't take place very rapidly. But, I think that we going to see some very interesting 4P blades and things like that. It will really start to gain speed.
We, of course, have a vested interest in pushing things in that direction because that is where the competition's lack of scale is going to hurt them.
There is a lot of stuff we are doing as well that will make 8 sockets, 16 sockets and 32 sockets much more affordable.
[Hear Richard discuss the benefits of the open Hypertransport spec here.]
El Reg: What OS would play well on those systems?
HR: I think both the Microsoft camp and Unix camp are eyeing that part of the market. I think Solaris, because of its history, has a lot of scalability and is promising. Sun will do everything they can to make sure it performs well on the large systems.
El Reg: When do you think poor old SGI will make the move to Opteron?
HR: My assessment is that any of the vendors who are still entirely dependent on Itanium must be very worried. It is one thing to have Itanium as part of a product line. It is another thing when your entire business hinges solely on the future of Itanium.
I think the future is dark enough or questionable enough that it would give certain companies great concern.
From that perspective, you would expect SGI to look at Opteron as either a viable diversification or, in a more drastic type of decision, a change in strategy. By not having Opteron today they are hurting themselves.
El Reg: You guys used to sweet talk Dell a bit in the press. Lately, I've noticed some harsher statements where you say Dell is missing out on a market. Have you given up on Dell?
HR: I wouldn't read too much into the comments we made months ago. I think we were reacting to something that we felt was unfair, which was statements by Dell executives that they were seeing no demand in their customer base for AMD systems.
I continue to maintain that is not true. I have meetings with large customers, and we know that they have asked Dell to reconsider their position.
That said, they have acquired Alienware, so they're technically a customer. We enjoy a very large share with Alienware.
There are no hard feelings between Dell as a corporation and AMD as a corporation.
El Reg: How far have the discussions with Dell gone?
HR: Our executive teams are talking to each other on a regular basis.
I have to be optimistic about our long term plans. Our long term plans are to capture 30 per cent of the business in both units and dollars. If you think that we have a chance to get there, I think that it's difficult to imagine a world where we have that sort of a share without some sort of a relationship with the number one player in the industry.
I think Dell's ambition is to get bigger, and AMD's ambition is to get bigger. So at some point in time, somewhere down in the future, I think it will be more likely that we have a relationship.
[Richard talks about the AMD and Apple deal that wasn't here.]
El Reg: Isn't it hard to keep banging the anti-trust drum given your success with the 64-bit chips and market share gains?
HR: I get this question a lot. It's a bit like asking, "You've been a criminal all your life but for the last three years you have behaved properly. Does that make you less of a criminal?"
El Reg: Yeah, but these lawsuits take a lot of time, energy, and money. Why bother at this point? Why not stay focused?
HR: The evidence is the evidence. The fact that we have been successful could be because we did the lawsuit. It could be because we started to raise the awareness in the consciousness of the customers, the government, and to a certain degree Intel employees that things needed to change.
We still see monopolistic behavior taking place every day. In some cases it might be more subdued, but that is not to say it has completely changed.
I am proud that Hector had the courage to say, "Enough". Somebody needed to do it. ®