Intel’s tri-gate announcement: too many angles for 350 words
We have just passed one month since Intel made its tri-gate announcement. In this time many things have been published about tri-gate; some insightful, some confusing and some, well just plain funny.
Moving beyond the initial flurry, and with plenty of time before tri-gate’s real impact hits, there are abundant avenues that deserve further study. There is a semiconductor process story, a design story and an intellectual property story. I think the best one can hope for in the remaining 240 514 words is to present a framework for digesting the above threads of information.
Marketing has always been something at which Intel excels. Historically they developed brand awareness for internal unseen components. People knew the Intel brand even though they would likely never touch one of their products. Is this a bit outdated or naive for technology people? Not really. Intel has now developed brand awareness specifically around process, making it sexy. Process has always been and always will be important to advances in the semiconductor universe, but it has rarely been covered by popular media. With tri-gate the interest has lead to a wave of comments and articles presenting an unholy confusion of process and design concepts.
The tri-gate announcement was a process announcement. There are no ifs, ands, ors or buts about this. It is Intel’s stake in the silicon for the 22 nm node. Will it be the only way of doing things at this node? No. Other foundries and integrated semiconductor houses will have their own solutions for 22 nm. Intel has simply laid down the process gauntlet. With Intel at the forefront of process development for the last decade or so this is no surprise.
The design side of this story is readily appreciated by considering the mobile space. Tri-gate does not in itself resolve Intel’s dilemma here. Much was published about how tri-gate will upset ARM’s prowess in mobile. Is that right? Is it that easy?
ARM is a vendor of RISC IP cores. Their licensees then integrate this core into their own designs to bring their individual flavour of and strengths in design to life. Nvidia leverages its experience in GPU cores and TI in DSP’s to name but two. Then there is Apple. Apple has decisively moved down their own design road with the A5 and there is probably no looking back. There is so much more on the A5 than the CPU that it is a shame to see it simply referred to as “ARM”. To say that tri-gate will pull this diverse set of design roadmaps over to Intel’s Atom or more generally x86 architecture seems naive. However, it may not be coincidence that the “Intel as Apple foundry” rumours started up around the same time. It will no doubt be a challenge for Intel to determine if and when it breaks its process lead away from its in-house design to capture margin from process.
The intellectual property story is of course interesting in its own right. Despite what some have written there is no one patent that Intel owns. Technology Review published an article that discussed FinFET’s roots and IP developments at Berkley. This was followed by a story from Berkley itself. A very simple search of the USPTO with Intel as the assignee and “tri-gate” in the specification found over 100 issued patents. There is no doubt that this oversimplified search caught some that are unrelated and missed some that are. But by scanning the titles it is clear there are plenty of developments that bring Intel’s version of FinFET to life disclosed in this set.
While it is still early in this story you can say that Intel is making process very strategic and may begin to more aggressively leverage their work here.