Intel Corp. is launching its 64-bit Itanium processors in the midst of a deep industry downturn and, as part of its effort, is seeking to rally support around its vision of 64-bit computing. At the same time, the leading manufacturer of microprocessors is moving forward with new mobile chips and is looking ahead to computing designs beyond todays CMOS devices.
Company CEO Craig Barrett discussed these and other issues, such as three-dimensional organic biological molecular compute elements, with eWEEK Executive Editor Stan Gibson following his remarks at Gartner Inc.s 2002 Symposium/ITxpo in Orlando, Fla., earlier this week.
eWEEK: Intel has placed a huge bet on Itanium. Can Intel succeed if Itanium does not hit a home run? Have you hedged your bet on other technologies?
Barrett: We are succeeding without Itanium revenue today. Were investing heavily in it, and its not contributing positively to the bottom line. Were succeeding today, so I assume we could succeed in the future. But it is a part of our overall road map. We make building blocks for clients, for servers and for networks. Itanium is the building block for servers.
eWEEK: Looking back at the rollout of Itanium and Itanium 2, is there anything you would do differently?
Barrett: We were a bit slow in getting the Itanium 1 product out. I think we did the right things, working in parallel with the OS and ISV communities in getting the product into the marketplace. We can always improve on our execution–we can do that on every product–but were very happy with the characteristics of Itanium and with the Itanium strategy.
eWEEK: Are the Itanium follow-ons, including Madison, Deerfield and Montecito, all on schedule, with Montecito arriving in 2005?
Barrett: Yes, theyre rolling out basically in a one-a-year timeframe.
eWEEK: Are you satisfied with Dell [Computer Corp.]s posture on Itanium so far?
Barrett: They were basically slow to come to the party. The comment they made about two weeks ago–which was that their customers are getting pulled to the Itanium 2 family–shows theyre moving in the right direction.
eWEEK: Is something called Yamhill [a 64-bit architecture chip thats compatible with x86] more than a rumor?
Barrett: I dont ever comment on rumors. Theres been a lot of speculation in the press about what Intel is doing. We have a 64-bit strategy, and it happens to be the Itanium processor family.
eWEEK: With regard to the LaGrande security initiative, in some quarters there has been noise of surrendering control. Is there any reason to that reaction?
Barrett: Anytime you do anything with security and privacy, theres a degree of concern. We tried to do something with processor serial numbers a few years ago. From a security standpoint it made a lot of sense. But if you can identify users on the basis of a serial number, then they may get upset.
Security is increasingly important. There is no questions that people will use PCs more and more for commerce, so you want to make them secure. What you want to do at the hardware level is facilitate what you want to do in software. In multimedia, we put multimedia extensions in the instruction set. Security is not that much different.
Page Two
: Barrett Interview”>
eWEEK: Banias is the code name of the mobile chip youll be rolling out early next year. Can you elaborate on how it supports wireless communications?
Barrett: Youll have support of dual-band 802.11a and 802.11b with the Banias architecture. The architecture consists not only of the processor but also the peripheral chips.
The platform of the future will have all the connectivity built-in. Bluetooth, wired and wireless, both wide-area network and local-area network.
eWEEK: The Tablet PC is getting a big rollout next month. Do you see that having an impact on chip sales?
Barrett: Its hard to tell. However, any innovation is good for the industry. If you ask me what I want built into a system, Id want wireless capability rather than handwriting recognition [as in the tablet]. But the tablet is a new usage paradigm, and usage models will vary with the individual.
eWEEK: Carbon nanotubes and quantum computing are the buzz for tomorrow. Do those technologies consume a certain percentage of your research and development?
Barrett: Were pursuing most of those activities in conjunction with universities. Its a relatively small part of our direct R&D budget. Those technologies are very much in the exploratory phase. The expensive part is the pre-commercialization phase, where it costs hundreds of millions of dollars.
eWEEK: With the cost of building a fabrication plant in the billions of dollars, are we reaching a cost limit for semiconductor makers?
Barrett: This is déjà vu all over again. Back in the 90s when we got to the first billion-dollar fab, everyone said we would have only five megafabs around the world. But then there were 150 more fabs built around the world. The issue is not so much what these things cost but whether you can get a return on your investment.
Its safe to say that fewer and fewer companies will be building their own factories because of the cost. Theres a lot of discussion about joint ventures and foundries. But we think there is a tight correlation between the design, the process technology and the manufacturing. We think controlling those gives us a performance advantage.
eWEEK: Are you a believer in grid computing? What impact will grid computing have in Intels future?
Barrett: Grid computing tries to harness unused computing power to make a more powerful computer. Weve had a huge over-investment in optical fiber capacity. Thats a way to hook up computer centers and make grid computing work.
But you go back to the local level, and grid computing has been around for a long time. We wire together 12,000 engineering workstations at Intel. That could be the largest supercomputer in the world.
Our industry is famous for rediscovering and repackaging things. The concept of distributed and parallel computing is not new. I believe in it–yes; were using it, yes. Is it real? Yes. Is it new? Its not that new.
eWEEK: Looking into the future, youve said that Moores Law, as we know it, will continue for 10 to 15 years. Are you familiar with Kurzweils Law, which says that three-dimensional molecular computing will grow at an even faster rate?
Barrett: Were getting three-dimensional atomic computing today.
Theres been a lot of talk about molecular computing and the impact it could have. Despite what the basic computing element is, you still have to interface and have communication with those compute elements.
Its the same challenge we have in wiring together five hundred million transistors on the next-generation Itanium processor. If youre going to have three-dimensional organic biological molecular compute elements, you still have to get information in and out.
So Im more concerned about the global wiring diagram to make it useful. You can shrink the transistor, but you still have to communicate with the transistor.