Intel has managed to keep pushing Moore’s Law by developing a 3-D transistor that allows the chipmaker to deliver ever smaller chips that will be more powerful, yet consume less energy. The race to build ever smaller chips to meet Moore’s Law ensures our electronics get more powerful and still decrease in costs every 18 months to two years. However, shrinking our chips has become more and more difficult and expensive, and people worried that we had reached a plateau. On Wednesday, Intel proved them wrong — for a few more years at least.
At a macro level, this announcement is big because it helps Intel keep moving down the process node and cramming more
chips transistors on a transistor chip in order to keep following Moore’s Law. And for Intel, the news is a breakthrough that could help it compete against ARM-based chips with their lower power envelope. Intel is fighting for its dominance with ARM as the mobile ecosystem ARM has reigned over becomes a larger slice of silicon sales, and as software improvements and concerns about power are giving ARM an opportunity in the server market.
The breakthrough comes from building up the channel that conducts electricity inside chips from a flat structure to a taller, 3-D structure. The stop-start flow of electrons through that channel is the underlying pulse of the zeros and ones that our computers read and translate into our web pages, documents and everything else.
The new design has already been manufactured by Intel and will be implemented within the next year to two at Intel’s fabs, although Intel did not detail the time frame. The first chips using this technology will be for servers and PC and laptops. Intel also plans to transition its Atom line of chips to this manufacturing process, which given the reduction in power could put today’s Atom chips in a similar power envelope to today’s ARM-based chips. And since Intel has the software edge on ARM as well as marketing might, this move, while not a checkmate, certainly could stop ARM’s momentum in the data center, laptops and even tablets.
Here are the basic stats about what this transition to 3-D transistors means for Intel and its chips:
- A 37 percent increase in performance, although this is comparing a chip manufactured using the 32 nanometer process to one made at a 22 nanometer process. For more on process nodes, check out this article. Mark Bohr, a senior research fellow with Intel, said the performance when compared across a 22 nanometer process would be more like a 20 percent gain.
- A 50 percent decrease is power usage. Again this isn’t a 22 nanometer to 22 nanometer comparison, and every time one shrinks the process node it yields power savings.
- A 3 percent increase in costs per wafer to manufacture the new types of transistors.
- “We can wave goodbye to planar transistors,” according to Bohr.
The engineers onstage at the press conference didn’t detail what contributed to the cost increase or say how much Intel would spend upgrading its factories, although one did explain that it was similar to a normal upgrade cycle and wouldn’t materially affect Intel’s costs. The next generation of Intel’s processors dubbed Ivy Bridge will contain this technology and we should hear more about them in the coming months.
For those who follow the manufacturing dramas inside the silicon industry (I know, there aren’t that many of us) the news is significant, but not as big of a breakthrough as I hoped for. Bohr said this breakthrough will take Intel down two generations of the process node, from 22 nanometers to 14 nanometers before Intel will hit some more engineering roadblocks. That’s really only three to four years away, which means this doesn’t extend Moore’s Law by all that much. However, it does buy time and perhaps by then, we’ll have had some bigger breakthroughs in quantum computing, probabilistic chips or any of the other experimental technologies I outlined in a GigaOM Pro report I did last year on pushing processors past Moore’s Law (subscription required). For more, check out the video below.