Intel is finally getting back to what led to it becoming the leader in electronics industry — re-inventing computing. The American chipmaker, who has lost most of its chip-manufacturing leadership to Taiwan Semiconductor Manufacturing Company, announced on Wednesday a new method to make more powerful chips, still using roughly the same size — by “stacking” them on top of each other.
To understand this better, lets first understand how Intel initially became the chip manufacturing powerhouse. Till date, Intel has used the principle behind ‘Moore’s Law’ — devised by Intel founder Gordon Moore — to manufacture chips. The law is more of an observation, wherein number of transistors double every two years, thus roughly doubling their performance.
However, in his own paper published in 1965, Gordon Moore had mentioned that this phenomenon would only last a few decades, which would then prompt a fresh thought on chip manufacturing. Those decades are mostly over since those transistors have shrunk to just a few nanometer apart. Intel itself has fallen years behind schedule on its own plans. Such is the lag, that the company’s newest 10-nanometer manufacturing technology will not arrive until the holiday shopping season of 2019.
Meanwhile TSMC has scrapped Intel of its title of making the tiniest chips, by introducing its latest chip manufacturing technology. The company is now already making chips for giants like Qualcomm, who have abandoned their manufacturing and outsourced the same to TSMC.
Intel needs re-invention, and it has set out on an ambitious plan to do so.
The company is now planning to get that tiniest chip title back, by stacking transistors on top of each other as layers. Intel said it now has technology to stack computing circuits on top of each other and plug them together with faster connections, enabling it to pack more onto a single chip. While the tech is already in use in memory chips, Intel will be the first to use the same in ‘logic’ chips or chips that perform computing tasks in a machine.
These new designs will be broken into something called ‘chiplets’. This will essentially allow Intel to provide consumers with a mix of both memory and computing chips. For example, a chiplet of memory chips and chiplet of logic chips could be stacked together to provide faster computing as well as ample storage, all by maintaining the same chip size.
As for availability, the technology will be available by the second half of next year.
We’ve been working on this packaging technology for nearly 20 years. There’s some real physics problems to solve in stacking logic on logic.said Raja Koduri, Intel’s chief of chip architecture in an interview to Reuters.