In the final instalment of ZDNet UK's Eye2Eye interview with AMD chief executive Jerry Sanders, he explains why he thinks Intel is abandoning the 32-bit market, and continues his views on the 64-bit arena. Editor Richard Barry also asks Sanders about the rumours that have shaped the microprocessor industry, and that spat with Gateway.
Jerry, you've spoken about Itanium and the way you believe Intel is abandoning the 32-bit market. You've talked about the golden opportunity for AMD as it stacks up against Willamette and the next generation of processors from Intel. So, can you say, categorically, that the processors you bring out in the 64-bit arena are going to be as fast and as well supported as Intel's? More importantly, can you promise that you're going to be able to ship them in volume?
Richard, I can't say, categorically, that I'm going to make it through the day! I can't categorically say that we are going to vanquish the largest company in the western world.
So how do you keep up with it?
Well, I think you have to look at it from the customer's point of view. The customer has a job to do. What the customer wants to do with 64-bit is address huge databases and other data-intensive applications. They want to have better memory capability. That's why they want to go to 64-bit. What they do not want to do is go to completely new software. Customers do not want to go to new software and new operating systems just because they're using a new chip. They will use the solution that best does the job for them.
Itanium is not that solution. Itanium, from what we have seen and heard, will run 32-bit software, at best, as an 800MHz solution. It has a large die, it's expensive and it's slow. We believe we have a solution with Sledgehammer that will better do the customers' job. We have talked to a few customers, revealed some details and they indicate we are on the right track. So we think we have a horse race here.
It sounds like you think Intel has made a real cock-up here.
I would never say that Intel made a real cock-up, Richard. Intel spent the money on advertising and turned the cock-up into Celeron!
What you have to understand is that the Itanium will come out first as a server. The people who want a server know the difference between 32-bit and 64-bit. The fact the Joe Public doesn't know 64-bit isn't that important. He thinks, "Jesus Christ, 64-bit... I thought Playstation was 64-bit!"
So what's the point? Well it's what application you are trying to serve? When the customer goes out, he says, "I'm looking to have X application running the entire management information system of my company." He knows what he is looking for.
There are benchmarks. We can show him a solution with Sledgehammer that gives him better benchmarks and better cross performance trade-offs.
Last week I interviewed David Potter from Symbian. He's the chairman of Psion (quote: PON). He clearly believes that Symbian is going to be the environment of choice for mobile devices, and I've yet to hear you talk of the next generation of computing. I asked Potter if ARM is his processor of choice, because Psion has a very close relationship with them. He mentioned a few other organisations and asked me to ask you whether AMD have any plans to move into that space in the future? Do you?
The answer to that question is "Yes". The answer to the next question is no further comment. Right now, what AMD needs to do is drive performance, because that's where the money is. The processors that are going to be used in the mobile space don't require incredible performance right now, but yes, we are working on a mobile solution.
Any thoughts on the Crusoe chip?
It's certainly a candidate for the mobile space, but I don't know what their strategy is right now.
Everyone in Asia, Taiwan to be specific, is going to use them for Web pads. I don't know what operating system they're going to run on that. It certainly isn't going to be an x86. And it's not going to be Microsoft Windows.
It's probably going to be mobile Linux.
That's a whole other game.
I just think that cellphones are going to win in this mobile war. I don't see the need for things like Palm Pilots. I really don't. When you get a cellphone that gives you the same functionality and you can make calls and surf with them... That's the future.
But don't you have plans long to enter that arena, in term of microprocessors?
I have to tell you that, right now, the focus has to be the movement to the server space and commercial space. You can't get enough volume to have a standalone business if you're only in the consumer space. You just can't get that.
Our ASP last quarter was somewhere in the $80 (£49) range. That's really not good enough. We've driven up our volume. Our original business model was to have a $100 (£62) ASP.
And the biggest challenge in that new arena?
Intel... and just the fact that they own everybody.
What about corporate perception of AMD? Put yourself in the position of an IT manager: There's this new AMD architecture for the server space, but he doesn't know it. Isn't that a significant issue for you to deal with? I mean, look at Intel a few years back with the Pentium bug... would you have been able to deal with a problem on that scale?
Well, I would not have put one out with a bug in to begin with. It put a bad chip out. All of those things are pure uncertainties. What if something goes wrong with some combination of 37 of our in-house custom programmes in concert with 49 commercially available applications at a low voltage on a hot day?
Oh my God! Yeah, you're right! That could happen. But that could happen to Intel, too, and it has. So at some point in time, a guy is going to say "how much performance do I get in this solution? What would it cost me?", but I think we'll win out. Otherwise, all of the value is captured by Intel. At the end of the day, that means no profit for anyone else.
This makes it very tough to take on established companies at the high-end. So our strategy has been "microprocessors are the masters". High-volume consumer applications are our entry point. I think we've done that. We've shown we've got great solutions.
I think that with all respect, we won with the K62. It isn't nearly as good a solution as the Athlon, which is a seventh generation processor, but we have a much better solution to come. K62 didn't make it into commercial applications generally because the market place didn't see it as a better idea. And I'm not going to sit here and tell you it was a better idea. It was a better idea for the consumer, because it did all the applications they cared about -- you know, all that floating point stuff.
But Athlon is a home run. It has a much better floating point than a Pentium III. It has a better performance and a superior clock speed, and that's scaring Intel. Y'know, you talk about motherboards. The motherboard guys are getting a lot of threats and scare tactics from the "gorilla". If you want supply of these hard-to-get chipsets, well let's think about it. You can't sell a motherboard without a chipset, right? If you haven't got a chipset to plug into that motherboard, you're dead meat!
At Intel they call the chipset that goes with the motherboard: they tell you how many chipsets you're going to get. If you want to do business with Intel, then my guess is that unless you're a pretty courageous guy, you don't make too many AMD motherboards. Or you do what the Taiwanese do, which is very clever, and say "OK, we just won't go public."
It's very, very tough.
Jerry, we were all expecting Gateway to launch with Athlon.
So were we.
What happened there?
Intel made them an offer they couldn't refuse. Think about that. So Gateway didn't launch with us and later regreted it. It wasn't that they were getting no chips, they just weren't getting what they were supposed to get.
But I can't add anything to that... obviously.
OK, so Intel will be using VLIW in its 64-bit plans. Is that part of Sledgehammer's plan also?
Absolutely not. It turns out we think their implementation of VILW isn't that good. Basically you're looking at dropping massive amounts of memory. That's what you're looking at.
Now we understand that there's another one behind it called McKinley, which was designed by HP, not by Intel. That's going to be better.
To be fair, most people are getting excited about McKinley rather than Merced.
Most people were excited about Merced, or so they thought. Right now you can be excited about a lot of this stuff, but we don't know what voltage they will be using, how much power it will be dissipating or what distribution it will use.
A lot has been written about the race to 1GHz. How important is it to get there first?
I think it is very important. The question is: Does it mean showing one or a dozen, or a thousand or tens of thousands? We're going to be producing hundreds of thousands of 1GHz processors in the second half of the year. Hundreds of thousands -- you know, significant numbers, although probably not millions.
Can we take it from this that you don't have dinner with Craig Barrett [Intel chief] very often?
I've been disappointed with him. We have to go back to when we won the lawsuit in 1994. We then had to sit down and carve out an agreement. We wanted to avoid litigation. Intel refused to give us a licence on what is called the P6 bus. That meant AMD would have to construct one entirely on its own. They deliberately set a barrier there, and I talked to Craig about that.
That was a deal breaker. They were not going to licence anybody on the P6. I didn't believe that for a minute. I knew he wasn't going to licence us. We were the only competition. So we carved out an agreement. And we started working.
At the time, I tried to impress upon Craig that this exclusionary stuff was anti-competitive and would not result in a good outcome for the user. He said they weren't trying to exclude us, they considered us legitimate competition. Legitimate competition -- we won the law suit, right? We won the arbitration. We were legitimate competition, not like some of the other guys, you know, laundering the technology. Craig's attitude was always: "We are just going to out-revenue you. You are not going to be able to keep up. You won't be able to keep up with the spending, you won't be able to keep up with the technology. We're just going to outrun you. Give it up."
Ha! I've always been... OK let's try and see how we do. I thought, well how are you going to outrun me? Yeah they are certainly going to outspend me, but thank God for the German government and some creative financing. We managed to build a second megafab, so we have two. They have five megafabs making processors. Two sevenths is 28 percent. So I said, "now I have the capacity to get to my 28 percent goal." I'm OK with two megafabs.
We've got the capacity to get to 28 or 30 percent if you want to round it off. So how are you going to outrun me, Mr Barrett? Are you going to outrun me on technology? Well, he hasn't been able to do that.
Intel doesn't want us there. Intel doesn't want competition. Nobody wants competition. It's hard to believe anybody is in favour of competition. Who is it who wants competition? Well it's the consumer who wants competition. It gives them more choice, better availability, new ideas. So we're the only competition left.
If our processor had the same bus structure as Intel, so that all we had to do was match our processor with theirs, it wouldn't be an issue.
I made that decision back in 1995, that we were going to become the rallying point for alternative platforms. The only thing we have in common is that we run the same operating system. That's what the customer wants. The customer doesn't want a new operating system -- other than Unix or Linux. It definitely doesn't want a new instruction set.
I think you undersold yourself. I think you said you were at 16 or 16.6 percent, so to get to between 28 to 30 percent by the end of 2001 is a lot of market share to ramp up to. How confident are you that if I ask this question in a year's time, you say "hey I've got it"?
Very. If we get the commercial market, we'll do it. It if don't, we won't. This is a do or die thing. We must get the commercial market. We'll do everything we can to get it. We'll get a better solution or chipsets or infrastructure partners. We can't get 30 percent of the consumer market, because even if we had 100 percent of the consumer market, we wouldn't get 30 percent marketshare. Meanwhile, we have the same marketshare as them [Intel] in Flash. We make a lot of money in Flash competing with Intel. In fact, between AMD and Fujitsu, I think we've got about 31 percent marketshare. Intel has 21 percent marketshare.
You come back in a year and ask me the question then, OK?
See also Eye2Eye with AMD's Jerry Sanders, Part I.