Three Minutes with Rambus CEO Harold Hughes

Rambus Inc., the memory chip technology developer, is looking to the future world of multimedia as the way for its XDR memory chips to find their way into more mainstream products.

The memory technology is currently used mainly in game consoles, but Japanese DRAM (dynamic RAM) maker Elpida Memory unveiled a speedy DRAM chip based on XDR technology, the 512-megabit, 4.8GHz XDR DRAM. The chips are aimed at game consoles, HDTVs (high-definition televisions), PCs, servers and workstations.

But one knock against the company is that while its products may be used in some high end applications, they have not trickled down into the mainstream because they cost too much compared to DDR2 (double date, second generation) DRAM and other technologies. Regardless, Rambus CEO Harold Hughes, sees 2008 as a breakout year for Rambus technology to enter higher volume production as it wins over some HDTV clients.

He said people should also be on the lookout for XDR technology to make some headway into cellular phones and other mobile devices.

Hughes was in Taipei for the Rambus Developer Forum, and discussed new products in an interview with the IDG News Service.

IDGNS: What are the major product areas Rambus will focus on in the next year for its XDR memory technology?

Harold Hughes: I think 2008 will see a few big [technology trends] moving forward in computers, HDTVs and the game players, and some of the neatest stuff will come out on cell phones. Those are four markets where Rambus is well positioned, so we hope to have a great year.

IDGNS: What is Rambus working on in the mobile market? Handsets, ultramobile PCs?

Hughes: Consumers want everything that they have on a desktop in one of these things (he holds up an iPhone) and obviously the bandwidth requirements are very, very high relative to what used to be the case for a simple little phone, and all that needs to be supplied at incredibly low power. Well, we've made what I think are some astonishing breakthroughs in low-power technology that are in many cases as much as 80 percent over existing technology. In applications where you're using full motion videos (TV on mobile phones) much lower power with full motion videos.

We're looking forward to making our mark in the low-power and mobile market. Our approach will be to use the bandwidth cores available with XDR, but to use our very low power interfaces to drive those, and to do that not only through our suppliers but also to look to some of the major players in the cell phone industry to bring them to market in 2008 or 2009. It most likely won't be XDR as we know it, but it might be derived from XDR. It has to be high bandwidth and extremely low power and it has to have low packaging cost.

IDGNS: Can you give an idea of what you mean by low power?

Hughes: We have introduced some very interesting low-power technology. We put out a [research paper] and we showed that we could deliver 2.2 Gbps at 100 milliwatts. Mind-bogglingly low, 2.2 milliwatts per gigabit per second. We could run it off two-double AA batteries. The joke we was that we would try to get it to run off of a potato. We passed 3,600 terabits of data through our interface with two-AA batteries, which is the equivalent of 100 DVDs.

IDGNS: Why do you think XDR fits the HDTV market?

Hughes: The XDR products are doing very well in the gaming products and we'll talk about the HDTV market. It's truly amazing the kind of compute requirements there, but it's also a consumer electronics product so it's got all the issues of cost and it's also the size, the electronic migration, the fans and stuff, so it's really bandwidth per device, and that's where XDR really stands out, in bandwidth per device. Such that with one or two XDR devices, you can get adequate to deal with the 5-10 Gbps of bandwidth these televisions are using. That's where working with companies like Elpida and other licensees we can have an adequate supply of XDR so we can do great things in the HDTV market.

IDGNS: What PC trends will help push more Rambus memory into PCs?

Hughes: The new generation of computers is going to have these multiple core processors, which are just driving incredibly attractive and interesting PCs. What's interesting is the nature of the cores. They are not only standard compute cores, but they are also graphics cores. For us, we're very happy on multiple levels. That's going to produce a greater need for memory bandwidth. In particular, the multi-core processors that have compute and graphics have conflicting needs insofar as the memory architecture is concerned. The computing requires capacity and graphics requires bandwidth. Rambus has a product called the XDR which offers multiple gigabytes of capacity and quite remarkable bandwidth, and in fact Elpida has productized a 4.8GHz part, so that's a pretty speedy product.

IDGNS: Rambus appears to be doing well in products at the top-end of the market, such as game consoles, but how is the company working to target the mainstream and lower-cost markets?

Hughes: Keep in mind that we believe that embedded in DDR2 is no small part of Rambus technology and that unfortunately is the subject of lots of litigation and there's no indication that's going to stop in the near future. I think that's always inevitable with our business model. I think that's almost inevitable given our business model, putting aside the litigation aspects of it. For a 400-person company to have impact and for IP to have value, you must solve problems well out into the future. Those problems are only going to be usable initially be a relatively small subset, but as that then starts to proliferate through more supply and costs come down, then you see those products trickle down to the $200 laptop.

I don't see that that model is going to change. We are working on memories that have truly staggering performance. Initially those will probably be actual parts incorporating as many of our features as possible, but by definition because of their performance there will be a limited number of applications that need it, but X number of years later, not only will those discrete parts become cheaper and likely migrate into higher volumes, but the ingredients in those parts are very likely to go into very low-performance, low-cost applications that are used widely.

IDGNS: What else do you see for 2008 that's big?

Hughes: Games will improve dramatically. The software development kits are continuing to improve such that I think games will be able to take advantage of more of the cycles. You know how they CAT scan a brain when its thinking? Well I suspect if they were to do the same thing with a cell processor on current games, it would show that a lot of those transistors aren't being turned on, especially when you're using an abstracted development code.

Ultimately [game developers] will learn how to directly address hardware more and more and I think you'll see those games coming out more in 2008 and they'll probably be spectacular, frankly. Of course, I've given up on games. Competing with your son is, well, it's so humiliating. You're getting killed in a game and you don't even know what's going on.