Intel announced today that its upcoming Ivy Bridge processing platform, which will be based on a 22-nm version of its second-generation Core (aka Sandy Bridge) microarchitecture, will also utilize a new transistor technology called Tri-Gate.
The company says that Tri-Gate transistors, the first to be truly three-dimensional, mark a major change in the way the industry has done things for 40 years, and could revolutionize it. Here's a quick glimpse at some of the most important facts and figures about Tri-Gate transistors, and what they will mean for PCs in 2011 and beyond.
1.) Tri-Gate explained. The Tri-Gate technology gets its name from the fact that transistors using it have conducting channels that are formed on all three sides—two on each side, one across the top—of a tall and narrow silicon fin that rises vertically from the silicon substrate. On a traditional two-dimensional, or "planar," transistor, the gate runs just across the top. But on the vertical fin, transistors can be packed closer together. This provides enough extra control to allow more transistor current to flow when the transistor is on, almost zero when it is off, and gives the transistor the ability to switch quickly between the two states. This maximizes both power usage and performance.
2.) Why? According to Intel, Tri-Gate was implemented because it would not have been possible to continue Moore's law at 22nm and below without a major transistor redesign. With Tri-Gate transistors, Intel claims to have extended Moore's law at least another two years.
3.) How small is it? A nanometer is one-billionth of an meter. That means that more than 6 million 22nm Tri-Gate transistors could be crammed into the period at the end of this sentence. By contrast, a human hair is approximately 100,000 nanometers wide.
4.) How much? (And how little?) Intel estimates that Tri-Gate transistors are 37 percent faster than those used in the current 32nm process and will effect an active power reduction of more than 50 percent, but will only add 2 to 3 percent to the cost of a finished wafer.
5.) Some upgrades required. Intel will need to make upgrades to its factories over 2011 and 2012 to get them ready for producing the large quantities of 22-nm chips necessary to drive its many devices for the near future. (To give you an idea of the scale, Intel's factories currently produce about five billion transistors every second—or 150 quadrillion per year.) The company says, however, that the actual changes being implemented will not be more significant than have been required for previous process improvements.
6.) How long will it last? Representing the latest "tick" in Intel's two-stage development cycle (the Sandy Bridge microarchitecture was the most recent "tock"), these Ivy Bridge innovations will be around in some form for at least the next two years. But Intel promises that the technology will be able to scale to its next production process, at 14-nm, so don't be surprised if it extends well beyond that.
7.) Tri-Gate is not new. Intel research scientists first invented the Tri-Gate in 2002, but it's taken them until now to get chips using it ready for high-volume production.
8.) Where will you see this technology? Intel says that Ivy Bridge–based processors are ideal for both servers and clients, the latter particularly in thin-and-light form factors (such as desktops and nettops), and that the technology is expected to scale to Intel's Atom line of CPUs as well, allowing for their usage in an even broader range of systems. But with such low power usage, smartphones, tablets, and other mobile devices would seem to be not just possible, but likely. (Unfortunately, Intel isn't yet saying when these will hit the market, although devices that don't necessarily require a carrier - such as tablets - will likely be first.)
9.) When will Ivy Bridge arrive? You can expect to see processors and devices using them by the end of 2011, with product shipping in early 2012.