Q&A: Cyrix's Rogers on the NetPC and the x86

PCDN: How is business going now?

It's going great right now, though we've had a couple of difficult quarters. We didn't do the work we needed to in terms of design wins in the past. We've got in the cycle, rolled a couple of customers over, and now we're seeing the benefits of that. Vobis is a long-term relationship; we have [PC vendor] Tulip which covers 18 countries. Of the UK off-the-page relationships we have Granville Technology.

The Cyrix chip is not identical to the Intel chip. You have to have a BIOS change, different wiring of a socket, get [PC vendors] to understand you have a solution that's going to work for them. Tell them they're doing the highest performance Windows PC on the market today.

What about the NetPC: you have talked about how you see your GX processor being ideal for low-cost systems?

Obviously that's one of the things that's happening today and that's generating a lot of momentum. The GX will run like a 133MHz Pentium with much faster graphics and SoundBlaster emulation. We think [NetPC] will take effect in the first half of next year. Intel's got their goon squad to see if they can change [PC vendors'] minds.

Do you consider that yourselves and AMD compete with Intel on a level playing field?

It's pretty difficult if someone owns 95 per cent of the market. You have to work extra hard, but to say it's not a level playing field - to whine about Intel - is a little overstatement.

Cyrix is at the high-end and AMD is at the low-end if you range from 75MHz to 200MHz.. We will have MMX in the January/February time-frame and we're pretty confident that in Q1 1997 we can match up to Intel's full production line. If we give AMD an opportunity they're going to take it.

You're the only mainstream PC CPU company that is fab-less. Is that a problem?

No, we're fortunate that we have a manufacturing partner like IBM. Some people talk about 'If you have your own facility maybe you have a cost advantage', but if you look at the total design cycle, we get in the market way ahead of the competition.

In which areas do you feel Cyrix needs to improve?

I don't think we have worked hard enough on partnerships. What we need to do is team up with companies who can help us bring complete solutions. Companies like Compaq and IBM four or five years ago were making 45 per cent margins; today it's 20 to 23 per cent. They have a lot less dollars and they'd rather spend them on marketing. Intel does their [marketing] by themselves.

We have showed that we can produce innovative products. Microsoft showed our product line at the end of August for the PC97 [NetPC] concept. It's a full-featured PC; the architecture is different and it provides you with a different price point for thin-client network access. We think there's a lot of opportunity and we're ahead of the competition. Everyone is working on it and we'd love to see alternative architectures at 20 per cent or more.

How big is the opportunity for NetPCs?

Well, we've got 100 per cent of the market because we've sold five or six computers! The question is: is there opportunity for volume in its own right or is it going to cannibalise PC sales? Our customers are saying it's going to open up the market. There's a lot of excitement: we'll just have to look at Q2 and Q3 [1997] sales and see.

Six months ago Intel weren't interested [in developing a NetPC-optimised processor] and they were very vocal about it. A couple of weeks ago they started changing their minds and began staffing it up. We've been working on it for three years. Maybe Intel can do enough to be only 18 months behind.

[A NetPC processor] has got to come in to get to a 200MHz $500 PC in two years. The issue is how do you expand the PC market. US families with over $50,000 income already have three PCs. There's a perceived demand to get online but [people] don't want to pay $1,500 to do that. So the opportunity is to make a system that lets you play games, get online, get entertainment [based on] a chipset that costs about $50. In Christmas 1997 we'll get the first feedback but we're optimistic. The OEMs are very cocky and confident.

If the NetPC takes off and the GX becomes the chip of choice would you rethink your production capacity situation?

No, I dream about that every night [laughs]. To be realistic, [with] MMX [instructions] that absolutely could happen.

The capacity situation is that the GX needs a 0.6 micron process. IBM has a significant amount of that capacity. There's always the opportunity to get a second source.

How do you see the x86 architecture developing?

We continue to diverge. The three of us [Intel, AMD and Cyrix] are getting further and further apart. You have to innovate if you're going to be a leader and we're closer to the Intel leading edge than we have ever been. At 200, 225MHz there's a great opportunity. We're at a comfort zone in the market where there's mass for the first time.

At some point there's going to be a debate: how many transistors can you add to keep the speed up at a viable cost. It's not clear you can have both. To get the MHz up do you stop at 14 stages or do you add another pipeline? Do you want to go to GHz on a chip that does everything or do you want a core x86 CPU with variants for different applications?

As the architecture splinters, users select different types of software and clock frequency becomes an unreliable indicator of performance, it seems more difficult than ever for end users to select the right chip...

I'm a technical person so I don't get caught up too much in the bullshit. When Windows 3.11 came it was a dog on a Pentium because it was developed on a 486; NT was designed on a Pentium. 16-bits or 32-bits is bullshit. It's to do with [the way software is written]. MMX will provide the opportunity for soft modems, enhanced graphics, reduced overall costs. I think the real issue is moving from a 16Kb to a 32Kb cache. A much bigger cache is more significant than the MMX.

At 1997 Q4, we expect the GX to be right in the middle of the entry-level computer market, probably competing with P55C [Intel's first MMX chip]. Systems will have DVD, USB, 1394, 3D. These systems will between sub-$1,000 and $1,500.