New ARM processor will enable machine learning in smartphones
Image: INQUIRER.net stock photo
Machine learning has become one of the biggest topics in technology circles of late, with competing companies each developing their own version of a virtual assistant. While current mobile devices are still unable to fully accommodate such technology, this will change with the new CPUs and GPUs from mobile processor designer ARM.
The company claims that the ARM Cortex-A75 has a 22 percent improvement in efficiency over its predecessor. The Cortex-A55 has the highest power efficiency in the mid-range CPU class, while the Mali-G72 graphics processing unit (GPU) is said to have a 25 percent improvement over the preceding G71, reports The Verge.
Article continues after this advertisementApart from being designed specifically for machine learning, these new processors will also improve performance for augmented and virtual reality applications.
This means that future smartphones may be able to run longer when used for mobile augmented reality and virtual reality. Virtual assistants will also be able to perform machine learning tasks without connecting to the internet.
Image: ARM
However, it should be noted that machine learning tasks is still better handled when connected to the cloud. The new processors, particularly the Mali-G72 GPU, will excel only at executing tasks already learned by the AI.
Article continues after this advertisementARM labels the Cortex-A75 and A55 as part of the new Dynamiq CPU product line. These processor designs will allow chip manufacturers like Qualcomm to have greater flexibility in making chips. For example, they can make chips that favor long battery life yet still capable of processing heavy loads when needed.
A “realistic time window” estimate by ARM places the release date of mobile devices equipped with chips from their Cortex-A75 and A55 designs to come out by the first quarter of 2018. However, ARM also notes that Chinese phone vendors could release devices powered by these new chip designs within the year. Alfred Bayle/JB
