Intel’s ‘most significant announcement of the year’ was, as expected, to unveil its 22nm chip manufacturing process, but it went a step further with a ‘3D’ design approach which will mark a huge shift in the semiconductor industry.
As Intel struggles with new markets, it has Samsung biting at its heels, but its latest news shows it can still shift the goalposts for the whole sector when it comes to its processes. The new tri-gate transistor design will be the basis of the 22nm generation, which will reduce cost and power and could finally make Atom a suitable platform for mass market mobile gadgets. The significant step forward on the manufacturing front has also intensified speculation that Intel will establish a foundry business, though this would involve becoming a major player in making ARM-based chips ‘ especially if it really does have a chance of winning an Apple A5 contract, as rumored.
The tri-gate technology was first detailed by Intel back in 2002 and will now form the basis of the 22nm mode. Intel also showed off the world’s first 22nm microprocessor, codenamed Ivy Bridge, which should be ready for mass production by the end of this year. The tri-gate transistor is a major shift because the chip industry has always used the flat, 2D ‘planar gate’ approach. Intel aims to replace this with a 3D ‘silicon fin’, which rises vertically from the substrate.
One effect of the design will be to allow chips and devices to switch very rapidly between ‘on’ and ‘off’ modes, and to consume virtually no power in the off state ‘ important in mobile and battery powered products. “Control of current is accomplished by implementing a gate on each of the three sides of the fin, two on each side and one across the top, rather than just one on top, as is the case with the 2D planar transistor,” said Intel on the analyst call. “The additional control enables as much transistor current flowing as possible when the transistor is in the ‘on’ state (for performance), and as close to zero as possible when it is in the ‘off’ state (to minimize power), and enables the transistor to switch very quickly between the two states (again, for performance)’.
The company says the 22nm tri-gate transistor performs up to 37% better at low voltage than the 32nm 2D designs, making it suited to chips targeted at mobile products. At the same performance, the new transistor claims to consume less than half the power of 2D 32nm chips.
‘This breakthrough will extend Intel’s lead even further over the rest of the semiconductor industry,’ Intel senior fellow Mark Bohr said in a statement. ‘The low voltage and low power benefits far exceed what we typically see from one process generation to the next.’ It will be vital for Intel to deliver on these promises if it is to make Atom truly appealing to tablet and mobile vendors ‘ a leapfrog, rather than the step improvements seen to date in Atom power consumption, is what is required, even if the new process is, initially at least, targeted mainly at the PC heartland.
As well as helping Intel to enhance its own processor strategy, its process breakthrough could signify a move into the foundry business. Intel has made some small but notable moves lately that suggest it wants to build a foundry business, making chips for third parties and creating a new revenue stream. In November, it announced a manufacturing agreement with start-up Achronix, in the area of FPGA programmable chips, and is rumored to have made a similar pact with another small player in this field, Tabula.
Now some analysts believe the firm is talking to Apple about replacing Samsung as its mobile processor maker. Apple is falling out with the Korean firm, having sued it for alleged patent infringement, and is said to be negotiating a new foundry deal for its A5 processor with TSMC of Taiwan. An Intel pact would likely kick in at a later stage since it would require the x86 giant to embrace an ARM-based platform, and prove it could match TSMC on price ‘ probably an impossible task with current processes.
Piper Jaffray analyst Gus Richard told EETimes that Intel’s manufacturing lead, especially when it moves to 22nm processes, would help Apple be more competitive with its home-designed silicon, and put clear water between itself and ‘Asian competitors that are knocking off its products.’ Even part of Apple’s business would be a major boost to Intel’s revenues (and a huge loss to Samsung, which has been closing in on the US giant’s chip market lead recently). ”Apple is the largest customer for semiconductors in the world. We estimate that Apple will account for roughly $20 billion of semiconductor demand this year,” Richard said. ”This represents 6% of overall semiconductor demand.’ Much of that, of course, is not in processors ‘ about 70% is for NAND flash and DRAM memory, plus advanced logic.
The 22nm progress could also help Intel reduce its reliance on TSMC, which manufactures some low power Atom models, though this deal is said to have run into some obstacles. Apple’s A5 currently uses Samsung’s 45nm process.