Last week, Nvidia CEO Jensen Huang hailed the death of Moore’s law at Computex 2017. This week IBM has resurrected it. Finding a way to squeeze 30 billion transistors into a fingernail-sized chip, IBM has defied prediction by coming up with a practical way to make 5-nanometer processors.
In collaboration with its partners Globalfoundries and Samsung, the trio of tech giants have created the technology using the same ultraviolet lithography (EUV) that they used for the breakthrough 7nm chip. Ditching the common FinFET transistor, in favour of silicon nanosheets, the change allows for an incredible number of transistors to be crammed into an incredibly small space.
IBM said the 5nm chips will be used to develop its own cognitive computing efforts as well as develop Internet of Things products. However, smartphones have not been given the IBM boot just yet with the chipmaker claiming that it imagines seeing phones with three times the battery life of current models.
Exciting right? Well hold on for a little bit longer, as IBM doesn’t expect to see any real examples of its 7nm chips until 2018 at the earliest. That means the 5nm chips probably wont be seen for at least another two years. However, it is still an exciting prospect for an industry that looked like it would have to rethink its blueprint just last week.
For half a century, Moore’s Law was taken as gospel. However, 52 years after Gordon Moore predicted that the number of transistors on integrated circuits would double every year (later changed to 18 months), and his theory was declared dead. Hammering the nail in the Moore’s Law coffin was Nvidia CEO Jensen Huang who took centre stage at Computex to signal the dawn of life after Moore’s Law. It was the perfect harmony for both consumer and producer, as computers twice as powerful as their predecessors would keep the consumer happy, while manufacturers enjoyed a built-in obsolescence.
The reality is, is that it couldn’t go on forever. But IBM is doing its bit to keep Moore’s Law alive for a couple more years at least.