AMD Goes on the Offensive
Despite a casual, low-key demeanor, Hector Ruiz knows something about a fight. During the last several years, the chairman and CEO of Advanced Micro Devices has been the public face of the companys crusade against its main chip rival, Intel. Besides fighting for market share and technological superiority, AMD is also waging a high-stakes campaign in the courts with an antitrust suit against Intel.
eWEEK Editorial Director Eric Lundquist and Staff Writer Scott Ferguson sat down with Ruiz in AMDs Austin, Texas, offices right before AMD was preparing to launch the much-anticipated four-core Opteron processors—code-named Barcelona.
It hasnt been an easy road to Barcelona for AMD: Ruiz readily admits that there were technical problems, and the companys financial outlook did not ease the growing pressure to respond to Intel, which has managed to bounce back in the last year with its much improved Core architecture. And, just a few weeks before the new chip was ready to come to market, Henri Richard, AMDs chief marketing and sales officer as well as one of the companys main spokespeople, announced his resignation.
The big news coming out of AMD right now is the launch of Barcelona. What can you tell us about this processor and what it means for the company as you move forward with your enterprise offerings?
Barcelona is part of our strategy that began with Opteron to become a significant player in the enterprise. When you look back at the early Opteron days, it was our first attempt to create a product that was really, truly built around the needs of the enterprise, and, being able to look back and be objective, it was quite a hell of a product.
AMD finally rolls out Barcelona. Click here to read more.
It changed a lot of the rules of the game. A lot of people have forgotten because it was 2003 when we introduced that product, and the introduction of that product really changed a lot of peoples view of where the enterprise was going. On one hand, you had the extension of 64-bit to 32-bit with x86, and that was really a smart thing to do.
In my view that was the beginning of the demise of [Intels] Itanium.
Back in those days, [Opteron] still really had a strong presence, but it did a lot of other things, too. It created an architecture that was very efficient.
At the same time we launched Opteron, we started working on Barcelona, which was the next step. The interim step that we took between Opteron and Barcelona was the dual-core introduction, which took the single-core Opteron and created a dual-core version of it, but all that time we were looking ahead to Barcelona.
People had invested in software as well as ecosystem—the hardware ecosystem and, of course, the software ecosystem. One of the things that dual-core Opteron taught us about was how powerful that was because people were able to upgrade to dual-core Opteron rather easily.
One of the things we are doing again with Barcelona is protecting the investment our customers had made for a long time. It is going to fit into the system, and its going to fit into the same thermal envelope, and, therefore, its going to have the opportunity for people to adopt it and use it at a much faster rate than they would otherwise.
The other thing that Opteron did was begin to change peoples views of energy-efficient computing, which was critical. So we believe that, in terms of performance per watt, [Barcelona] will continue to set a strong leadership position for our customers.
The other piece—again, we were looking ahead developing the Opteron family—was that virtualization was really going to become more and more significant over time. We put a lot of work into Barcelona to make it very efficient as a virtualization machine. I believe, and our customers are telling us, that in terms of performance relative to virtualization, it is going to be a phenomenal machine—very scalable and a lot of things that are very critical.
Then when you come down to the last piece, we put a lot of effort into putting in a world-class floating point capability in this device, so it will have very strong raw performance. So we are very excited and champing at the bit to see Barcelona out there.
What is the most significant technological advantage that Barcelona has over the dual-core Opteron besides the addition of two more cores to the silicon?
That is a tough question. I really think that when we look at Opteron and then when we look at Barcelona, Barcelona is an improvement on all the things we just talked about.
Page 2: AMD Goes on the Offensive
600 Million Transistors on
Was there any frustration on the companys part, or on your part, that Barcelona did not come to the market sooner?
There is nothing that we would have been more excited about than getting it out earlier. But we are not making excuses. This is a damn difficult thing to do, as Im sure you can imagine. This is the first time that any company in the world has put 600 million transistors on a chip—and you can round that off and say its a billion.
This is 600 million transistors on a chip, four cores, complex technology and tremendous architectural features.
If you could do it over again, would you have taken the approach to quad cores that Intel took with its Xeon processor, which basically ties two dual-core chips on the same silicon?
This is one of those things that at some point in time you have to bite that bullet integrating that much technology on a chip.
We felt that the quad core was the appropriate time to do that, and Im glad we did, even though we were disappointed that we couldnt be a little earlier. But now we have that experience, which is amazing.
We now have a talented technical team that has been able to figure out 600 transistors on a chip, and we have a talented manufacturing team that [worked to make sure] that when you had something this complex, that all 600 million transistors worked, and they have done a phenomenal job.
Read here about AMDs current road map.
We have gone through a lot of learning, and, as we have publicly stated, the defect density for our quad-core technology is the lowest we have ever experienced. In all, people worry that we are not quite at the same node technically as our competition. The answer we give is that, yes, that is true, but we [made] a conscious decision to be earlier at other things, and we were earlier in being able to integrate this much technology [on a chip].
We have the HyperTransport technology with this chip that is pretty powerful and the architecture features, like the floating-point capabilities. [If you look at it] from that dimension, you can say we are six months to two years ahead. There is no question that in our minds, based on the data that we have from our partners, that this is the best virtualization machine that is out there.
What about the fact that the clock speed that Barcelona will debut at—2GHz—is a lower clock speed than some customers and industry experts had expected?
Being able to build this product was incredibly challenging. So, being able to benefit from the scaling of the frequency was also a big challenge. We made a decision, and we recognize that the part we can claim the strongest position in is performance per watt. So we made a decision that we were going to be out in the watt range—that really put us in a clear leadership position.
This just happens to translate to a frequency of 2GHz. Frankly, I think we are going to surprise people. This is an area that we made a decision to come out in a place where we had the clearest and strongest position. I would predict to you that at the end of the year, people are going to be very pleasantly surprised in how well this is going to scale in frequency.
Any estimates to where its going to scale by the end of the year?
No. We are not ready to talk about that.
Page 3: AMD Goes on the Offensive
Responding to Intel
How has the market changed in the last year, and how is AMD preparing to respond now that Intel is on better footing than when the dual-core Opteron came out?
In some masochistic way, I like to take credit for Intel being a better company. I believe that is what competition does. We do recognize and we do acknowledge that we are going into a period of time quite different than the one we went into when we launched [the first] Opteron, and we are now in a much more competitive space.
[Intel has] good products. Before, they did not have products as good, and now this is going to make us stronger in a certain sense. Now we have to figure out how to do things even faster and better.
How do you intend to move the products out to market faster to improve your manufacturing capabilities?
I think we are doing some things that are going to put us in a strong position, but I would put [them] in the category of highly proprietary. I think that you can measure the results when you put [Barcelona] under the microscope.
I think you are going to see some highly innovative things, and that is only beginning. I think that our approach to multicore technology is different from our competitors. I think that it is going to make us a better developer of technology, and despite the size differential, by any measure our manufacturing is pretty strong. The fact that we are only one-tenth the size [of Intel] is a disadvantage to some people, but I believe that we are fairly strong.
Earlier this year, you lost some market share. How did your relationship with Dell affect the supply chain in the first half of this year?
You know that had a psychological effect on us, but it was unfortunately one of those things like the stock market, like the behavior of subprime lending. There were customers around the world that were concerned that the energy that would be required to serve such a large customer would take away from our attention to others. There were some people who were nervous and held back from buying stuff from us. But the reality is, when you look at Dells ability to ramp, in addition to addressing their own set of issues, that was more of a psychological reaction and an emotional reaction and not a real reaction.
We always knew what it would take to serve a complex and challenging customer like Dell, and, while we were prepared for it, the outside world did not anticipate it. Therefore, I think we suffered some unintended consequences for a couple of quarters. People thought that we were going to have to divert our attention, but I think that has settled down. Now, I think everyone knows that our relationship with Dell was orderly, and I believe that is behind us.
AMDs market share rebounded in the second quarter. Read how.
How does the new relationship between Sun Microsystems, which had been an exclusive partner of yours for several years, and Intel affect AMD?
Sun is a true partner, and we have done a lot of things in a joint fashion.
The relationship between our engineering teams is very strong, and, at the risk of sounding flippant, we really welcome the competition with Intel. I really believe that anyone that deals with both of us eventually knows that we have the better products, and if you dont deal with both of us, you will never know.
I can tell you in some way that this is good for us, and all the products that Sun is launching or developing continue to be AMD-based products. And Barcelona is going to be a very key part of their offerings.
Sun recently announced its eight-core Niagara 2 processor, which seemed to leapfrog ahead of the four-core offerings from Intel and AMD. Can you explain your views on what Sun has done for microprocessors?
Sun has had a very respectable computer architecture team that continues to enhance SPARC. It is very similar to IBM, which has a very competent architecture team that continues to enhance Power. Both of these companies have a strategy that continues to nurture their proprietary architecture, but I think both companies know how important a standard product is for them. I think our role is not to challenge that strategy but to serve them when they need x86 processors, and thats our plan.
Page 4: AMD Goes on the Offensive
AMDs Financial Picture
AMD has reported losses in the last few financial quarters. What are you trying to do to rectify the companys financial picture? Also, in the years between the release of the first Opteron and now, AMD grew very fast. Did you grow too fast for your own good?
We do recognize that we have a financial challenge.
In addition to all the product and technological challenges in front of us that we have to address, you have to start at the strategy of what we want to accomplish and the things that we want to do. We want our server architecture to continue after Barcelona and “Bulldozer” [a new set of x86 processor cores that AMD is designing and will debut in 2009]. We have ideas, through the acquisition of ATI, on the future of computing, and we have those plans in place.
You saw us improve from first to second quarter [of 2007] financially. It is our expectation that we will telegraph this to the industry and that we will again improve in the third quarter, and we will continuously improve. All of that is being done prudently and judiciously. We are not out slashing costs just to get a quarter with a right number because we truly believe in our vision. We believe in our products and our road maps.
The other question that you have is interesting—did we grow too fast? Quite the contrary. I think in any industry and in any environment, whether its automotive or airplanes or computers, when you have the technology advantage that we had from 2003 to 2006, we should have grown faster. It was only due to the abusive, monopolistic behavior of our competitor that we didnt.
Five years from now, what does computing look like? Is it going to be desktops and laptops, or is there going to be a whole new model out there?
We think visualization, not virtualization, is going to be the key to everything. The other part that I think is going to be big is the segmentation of the market, to make it easier not to have a general-purpose machine that does everything for everybody but to be able to adjust. This is where our multicore technology is pretty strong, and we will be able to do things that are great for supercomputers and, with minor changes, for servers and workstations. With further changes, [the multicore technology] could be even better for desktops.
I also think that desktops are going to become an appliance. Today, it is pretty easy to say that if you have a four-bedroom house, a three-bedroom house, you have a phone in each bedroom. I think the desktop will become an appliance of that type.
I think that mobile computing will continue to evolve pretty rapidly, and there are going to be some accelerated conversions as to whether you want a laptop that does a lot of things or a phone that does a lot of things. I think there is going to be a mixture of products in there, but I think definitely mobility is going to be key.
What can you tell us about the upcoming quad-core Phenom processor for desktops, especially in light of AMDs recent success in selling desktop chips this year?
As far as desktops go, in the foreseeable future, [this part of the market has got] two big buckets.
We play, and our competition plays, in the performance-driven bucket. That is, if you are a gamer, and you want great technology, that is where Phenom comes in. It is a derivative of the server processor—Barcelona. Thats what people saw at our analyst meeting. Its pretty amazing that it was able to demonstrate some gaming capability that is not available yet in the market because of the technology. So we believe that we are going to have a very strong position in that space.
The other space is the very price-sensitive, appliancelike space. I think that is more about features and platforms. There is closer interaction with the OEMs to figure out what kind of platform you would like to put into it.
Mobile is going to be much more platform-driven than anything else. If you think of the phone as an example, when was the last time you bought a phone because it has a 1GHz processor in it? I would guarantee that you wouldnt know what the speed of the chip in your phone is. You care about what you want [the phone] to do.
What we see more than anything else is technology moving fast to stores to provide the experience the user needs. There are some people who want to play games on mobile [devices], and that means that they are going to want to have certain capabilities on that device. This is quite different from the homemaker who might want to use the product as a Web surfing technology, for e-mail, etc. So we need to be able to address these issues and answer them.
This is something my competitor hates to hear, but I think, frankly, that the CPU is going to be less and less relevant. What I think is going to be relevant is the platform.