Intel is on a seemingly unstoppable march toward manufacturing unimaginably small chip geometries of 15 nanometers and beyond. Generally, the smaller the chip geometries--from 45 nanometers to 32 nanometers, for example--the better the performance and the lower the relative power consumption. And to achieve smaller geometries, the chip giant announced today.
I talked today with Brian Krzanich, a senior vice president and general manager for manufacturing and supply chain at Intel, about the future of Intel manufacturing and, consequently, the future of its processors.
Note that Intel is slated to bring out the 22-nanometer Ivy Bridge processor family in the second half of next year. Sandy Bridge is due to appear in systems during the first quarter of 2011.
Q: What does the Oregon investment mean?
Krzanich: The development fab up in Oregon is really targeted at 15 nanometers and beyond. And it doesn't even get completed until 2013. So, that's really going to look at 15 nanometers, 11 nanometers, and 8 nanometers and beyond. The real innovation and value is bringing these processors to market in high volume very quickly before everybody else.
So, that development fab is an example of the investment that we're making to make sure that continues. It was 32 nanometers two years ago and 45 nanometers two years before that. Every two years, we're going to make those investments. But when you invest in a development fab (like Oregon), it's means that you believe in an even longer term future. That we can keep doing this.
You're moving to 22 nanometers with the future Ivy Bridge chip. What about 15 nanometers, is that real?
Krzanich: It's absolutely in the labs. And it's next on the dock after this 22-nanometer ramp. Ivy Bridge (22 nanometers) will be the follow-on to Sandy Bridge. Ivy Bridge will not only have a smaller footprint. It will take Moore's Law and add more transistors, which means more features and generally you have lower power. Ivy Bridge is like a second generation Sandy Bridge.
What is Intel focusing on now? Aren't you beginning to gravitate away more from traditional CPU (central processing unit) to other functions?
Krzanich: It's no longer how just fast you can do the quarter mile. Now, by integrating memory controllers and graphics you're going to see more and more memory on the parts...additional micro engines that can do things on the CPU. That's really the future. It's still pretty fast in that quarter mile, but you're getting more performance now because the graphics is on board and coupled (with the main processor).
The memory is shared and routed on chip. It's like going from a drag racer to a Porsche, not only does it go fast but it can turn quickly and sticks to the road. It's a more balanced view of the platform. The (traditional) CPU is still the center of the universe--we're still adding cores...we've talked about eight, ten, and twelve cores--(but) it's not all about the CPU as much anymore.