Nearly every smartphone in the world is powered by an ARM chip and now, the company owned by SoftBank, have announced new micro-architecture that will push the demands of Artificial Intelligence.
This new chip will be built using DynamiQ, a technology that will allow manufacturers to connect a wide variety of CPUs allowing for more powerful systems-on-chip as well as processors that serve the A.I of the future from simple computing tasks to self driving cars. John Ronco, ARM product marketing head, explained “It’s a step change in how we build CPUs and the way we stitch CPUs together. It’ll be in smartphones and tablets, for sure, but also automotive networking and a whole range of other embedded devices. Anywhere a Cortex processor is used today, DynamiQ is going to be the next step forward.”
With ARM’s new CPU architecture, multiple processing cores will be clustered together with one that is tailored for the right software including one that can efficiently handle AI algorithms. Chipmakers will be able to now develop CPUs with up to eight cores and new software libraries will be released for the most popular AI techniques to run on ARM’s processors. Over the next three to five years, the company has claimed that this new technology will boost AI performance up to 50 times more than those currently available.
Ronco announced that he expects smartphones with the DynamiQ architecture will be available on the market as early as 2018. This is a huge shift from ARM’s previous approach big.LITTLE which was first introduced in 2011. This design contains two sets of big and small core processors to power the application and reduce battery consumption and has been applied to chips in all major smartphones including the Qualcomm Snapdragon processors for Android devices and in Apple’s iPhone chip, the A10 Fusion.
Not only will DynamiQ offer additional flexibility for chip makers, it will also lets them optimize silicon for machine learning tasks. There will now be the option to build AI accelerators directly into chips so that systems can manage memory and data more efficiently.
