Rambus CEO looks at future operations, deals

Rambus veterans Dave Mooring and Subodh Toprani took over as president and head of the newly formed ventures group, respectively. The company also announced it would double the data transfer rate and quadruple the bandwidth of its Rambus DRAM technology and apply that to new areas, such as gaming and communications.

It's been a rocky year for the company. Rambus' seemingly bright future darkened in September when Intel unexpectedly delayed its 820 chipset, also known as Camino. The delay, the second, stunned many PC makers who had been counting on Camino to enable next-generation Rambus memory.

CNET News.com spoke with Rambus CEO Geoff Tate about the company's plans as PC makers finally ship RDRAM-based systems after long delays.

CNET News.com: Could you explain what's behind today's organizational changes?
The organizational changes don't mean nearly as much as some people are making of it. Dave, Subodh and I have all been with the company for a heck of a long time. We're changing some titles, but all of them have been important people for us and we've all been running the company. I wouldn't read into it any important change in strategy.

As we are getting bigger and taking on more things, we decided to split my job into two, with Dave and I running the company as two in a box. We're just starting to do enough new things that it was more than one person could handle. He's going to focus on the operations or running the business units. I'm going to focus on strategy and intellectual property matters.

So what are some of these non-operational areas you will be working on?
One thing we've been doing is increasing our efforts in our intellectual property area. We now have a staff of five, up from one person a year ago. We've always been active in the area of patents--today we have more than 70 issued patents--but we are increasing the rate of filing new patents. And that's really just taking care of the increased innovation in the company.

There's talk double data rate (DDR ) memory may violate some of your patents; is that correct?
We really don't know for sure until we see some chips, and we've now gotten some DDR chips. We plan to do a thorough analysis and release the results of our analysis during [the first quarter], so we can't comment today. DDR, by our definition, is not a real competitor to Rambus, but it still may be using some of our IP [intellectual property].

Could you explain what your new ventures operation will be doing?
Subodh Toprani will report directly to me and will start up the new ventures group--and that is starting from scratch. We have never acquired any company in the past, and our expectation is we will do so in the future. Those companies would be companies that would be run by our president Dave Mooring. The idea is not for Subodh to identify companies and run them, but to go looking for companies that would achieve the goals of our strategies.

Are we talking IP companies or are you thinking of moving into hardware?
We like the IP model. Our best guess on where we will do an acquisition: We're talking about applying our chip technology to new applications. We have some great technology for transferring data at very high speeds between chips, and that technology is going to keep getting faster. We applied it to the memory market because it just made sense, but we see a lot of other bottlenecks in systems. And our customers have asked us to help them in new areas, and we?ve applied our technology to at least one new application.

And that application is?
That's something we will announce next year. Today we're just talking strategic goals. Next year there is going to be a lot of product and customer announcements that wouldn't make a lot of sense given our prior strategy, unless we announced our expanded strategy.

So what kind of acquisitions are you considering?
Our first acquisitions will be companies that have know-how in different application areas where we can use their application know-how with our chip connection technology and get a system solution to solve that system bottleneck more quickly than if we tried to figure it out ourselves.

Such as networking?
That's a possible area. Our focus is computers, consumers and communications, and that covers just about everything. We're going to be working with the same kind of companies that are using our memory interface technology and helping them solve other problems in their systems as well as memory interface technology. We're just selling them a new solution for a different problem in the same box.

What are some of the bottlenecks you see in PC systems?
The kind of bottlenecks related to our technology we hear about are I/O buses and front-side speeds, which is a function of, among others, signaling technology. There are also inter-processor buses. As things go more to multi-processor systems, the more quickly you need to move data between processors. So those are all potential applications for our technology.

Some of those areas are very contentious right now, such as competing I/O standards.
We aren't by any means saying that we overnight are going to magically apply our technology to every problem in the world. We're going to pick one or two bottlenecks and do a very job at them like we did in the memory systems area.

You're planning to significantly improve Rambus memory. When can we expect to see that?
Our expectation is that in the first half of next year we will demonstrate publicly and detail exactly how it works. We have this technology working in our lab today. We also recently showed it under [non-disclosure agreement ] to some of our partners and we will disclose the details under NDA over the next couple of months. The actual technology transfer will take place later in 2000, and by then the actual productization of it and where it will appear should be clear.

So what kind of customers do you expect to adopt the newer technology?
Our best guess is that demand will be from customers that have high performance needs out of a single DRAM. If you have just one DRAM and you need more performance, the only way to get it is to have the wires of that DRAM work harder and harder. So counter-intuitively, the adoption of our highest-speed signaling may be in small systems but once happen to have high performance, like consumer multimedia, graphics or communications products.

Do you just mean high-end when you say high performance?
In the memory interface phase, there's everything from $100,000 servers to $100 toys, and they all have different needs. We expect the vast bulk of the market will be satisfied by the existing products we just started shipping. But there's always the leading-edge high-performance guys who absolutely want the fastest they can possible get, and those will be the first adopters.

Interestingly, that's not always the highest-end computers. If you look at who first started using our technology in volume [it] was a toy company, Nintendo. That's because they had to get a lot of performance at low cost out of one DRAM.