CNET también está disponible en español.

Ir a español

Don't show this again

Tech Industry

AMD's CTO says Intel messed up

With a few strong years of market gains and product reviews, AMD isn't about to rest on its laurels.

There used to be an easy way to tell time in the processor business: Advanced Micro Devices would have three bad years and one good year, and Intel would have one bad year and three good years.

Strategic changes and new products like the Opteron, however, have transformed AMD from a company too often known for gaffes and financial losses into a major supplier of processors. Intel, meanwhile, has tripped over itself several times during the past three years.

Intel is now in a position to reverse its mistakes with a new line of chips. AMD's Chief Technology Officer Phil Hester, however, says that the smaller competitor isn't resting on its laurels. It continues to improve its chips and, with the acquisition of ATI, will release integrated chips for notebooks and other devices.

Will it be enough to keep a reinvigorated Intel at bay? That will be one of the big topics of discussion at the Intel Developer Forum next week in San Francisco. Hester, who worked at IBM for 23 years, says AMD is ready. He recently sat down with CNET News.com to share some of his views.

Q: It's kind of an interesting twist that AMD is doing "platformization"--that is, offering reference designs and more than one part to PC makers--because for years, the idea was to concentrate on processors and let your partners make chipsets.
Phil Hester: A lot of it is driven by what the OEMs (original equipment manufacturers) have asked us to do. It started off actually in the mobile space. Last year, for the first time, we did a reference design in the notebook space.

The OEMs would like us to do more of the validation and certification work. Each of them still wants their own unique packaging in the notebook space, but kind of a core design, if you will, that's been validated. So we work with two or three wireless vendors, a couple of graphics vendors, and then make sure that we pick selected elements from that.

But they don't like the Intel approach of saying, "You can only do this one, this one and this one." They want us to be able to say, OK, we validated with these--let's say, three wireless chipset vendors and these two graphics vendors.

There's been a lot of learning inside of AMD, both at the technical level and also at the management level.

How come you didn't do it five years ago?

Hester: This really came a lot out of what we call the commercial stable image platform business. In the commercial marketplace, they want a more rugged, robust platform than on the consumer side of things. And it really grew out of some of our success in the commercial side, which was driven by Opteron. Before that, to be honest, we were not taken seriously in the commercial side.

AMD has always wanted to get into the enterprise with notebooks and desktops. Why has it been tougher than it was with Opteron?

Hester: It's a work in progress. I think you've seen some major OEM announcements about going into the , and there are others in the works. Watch this space. It's clearly a focus for us, but it's an area where historically we were not stronger.

Are there any OEMs waiting in the wings with AMD PCs for the Fortune 500 crowd?

Hester: I think there are a couple that would be recognized worldwide, tier-1 names that have products under development. I can't obviously be too specific, you know, there are at least two that are well-known names that would have products.

The history of AMD and Intel is interesting. For a long time, AMD was the company that could do no right, and Intel would always be breathing down your neck when AMD stumbled. Why is it no longer the case?

Hester: I think several things. There's been a lot of learning inside of AMD, both at the technical level and also at the management level. (CEO) Hector Ruiz did a good job of reinvigorating the executive team at AMD and got it to be a more predictable company that the tier-1 (computer makers) can trust as a strategic partner.

The other half, of course, was that we have the product from the technical team. The Opteron got us in the door. You can show a slide where you say why you're going to be a great company, but until you've actually got a real product, people are always going to be suspicious.

Are you going to do platformization for commercial desktops?

Hester: Yes, we expect to.

With a variety of different chipset partners or with your new friend ATI?

Phil Hester: Same platform choices as before. We have no intent of trying to exclude anybody. Obviously the relationship with ATI will be much stronger, but we would have no intent of locking anybody out if they wanted to provide solutions to our platforms.

Are you guys talking to Apple Computer?

Hester: We have.

Has it shown any interest?

Hester: I'd say interest, not necessarily any decisions.

What are your longer-term aspirations in the server market? Is there any thought of doing a high-end, x86 chip that includes reliability features and capability features that a regular Opteron or Athlon may not have?

Hester: In general, the features that we add have played well in the server space and client space. There's more benefit in the server space, but it's actually more difficult to have two designs. It's easier just to have a single design that has all the high-end features in it, because you only have to validate one thing.

Are you looking at more than eight-processor machines?

This time next year, we will have a quad-core (chip), so an eight-processor system will be a 32-processor system. Probably 99.5 percent of all the servers in the world could be met by a machine of that capacity. We know how to do a glueless eight-socket system. We think, for at least the next couple of years, that is where the opportunity is.

If you look at the AMD-Intel race in the last few years, two factors played a big part: the changes at AMD, and Intel made a lot of mistakes--it did not, let's say, have the best chip in the world. Now it has a new family of chips that are getting good benchmarks. Is it going to get tougher for AMD to grow in the next couple of years?

Hester: In a healthy, fair and open environment with two strong competitors, you ought to be at a point where the products really compete with each other.

We had a period of time where it (Intel) just screwed up. I mean, I don't know how else to say it. It finally has responded to what we did roughly two or three years ago, but we've got our next generation coming out.

As long as there's a fair and open competition, it's a horse race between the two companies. And when we talked to the tier-1 (customers), that's what they want.

When I talk to customers, there are usually four or five different criteria that they use. Price is certainly one of them; performance is another one. What sort of innovation they can do in the platform is another.

Dell and IBM are probably good examples of companies that you know for sure are both our future roadmaps and Intel's, and they both chose to go with us for a broader and broader set of products. So, the way I read all that is we're doing the right thing across the board, right. It's not just performance; it's not just pricing.

Intel is going to continue to be aggressive. We're not at all confused about that, but we think we've got the right design points.

Another thing that comes up a lot is that AMD is talking a lot about incremental improvements for the basic architecture. Some people say that Intel plans to change architectures much faster and that AMD could fall behind technologically.

Hester: So what Intel actually is doing in my opinion is fixing a screw-up. If you look at the microarchitecture of the Core stuff, it looks a lot more like Pentium III. NetBurst (the underlying architecture of the Pentium 4) went off the deep end in terms of deep pipelining. It's a bad machine organization.

So they've had to go back and fix things that we never broke. There is nothing fundamentally broke with our core, and Intel is in a different position with NetBurst. That core was fundamentally broken, so (Intel) had to go fundamentally change it. And we never really went down that path. We went down a different path and said there's a balance between clock rate and parallelism.

There are certainly incremental changes you can make in the core, but the core itself is generally correct

With that in mind, how does the ATI acquisition contribute to that?

What's very interesting to us about the ATI acquisition is that now we can really make the right set of optimizations without artificial boundaries. There's a war right now between the CPU (central processing unit) and the GPU (graphics processing unit) guys, and to be honest, it's a bit of an artificial war. So a better approach is to be able to handle that stuff together where you can make the right trade-offs. If we need to move silicon between the CPU and the GPU, we can do that now.

Are there changes that you're planning to make to the core for the mobile space?

Hester: So, one of the areas that we need to work on as a company is the mobile space. And that's where the biggest win comes from being able to integrate the graphics.

Integration in the microprocessor itself or integration in the chipset?

Hester: Integration of the CPU and the GPU. Assuming the transaction closes on time, we would target a merged design in the 45-nanometer time frame.

Which is 2008?

Hester: Yeah. Some of the other things that are happening in the graphics space are that there's more and more programmability. It used to be that it was just polygon rendering. That's what graphics was, but now developers are doing so much programming.

The next generation of gaming is really making things more dynamic. It's not making the surface look realistic, but making it behave realistically. We've crossed the point where the GPU can do real programs of a significant size.

It may seem like 2008 is a long way away, but that's actually a major design cycle. ATI also has very good business, in both the handset and set-top box DTV area.

Are you looking at x86 in phones?

Hester: Yeah, absolutely. In high-end, high-function phones, it's driven by software. If you look at what's out there today, there's a number of embedded RISC (reduced instruction set computing) processors that are all fine. But as the software stack gets more complex, people want potentially to run the same applications.

Cell phones have been such a tough sell though. Intel landed the deal for the latest BlackBerry, but it's a rarity.

Hester: Again, if you go look at the processor architectures that are out there, there's a whole host of them. There's Power PC, Hitachi. What that says to me is today is not the right time to try to bring in an all-purpose architecture to that space. It's not needed yet. But as the software stack gets more and more complex, the software development environment becomes a bigger deal. Then the x86 is a good match.  

CNET News.com's Stephen Shankland contributed to this report.