ARM announces new processors; made to power the machine learning

ARM has been supporting the development of mobile devices since 1990 through processors. These processors are the exact platform on which companies are now developing the abilities to boost mobile AI, mobile VR, and machine-learning tasks.

Further promoting the development, ARM’s latest processors are built for machine learning and artificial intelligence. ARM has unveiled its next-generation CPU cores, the Cortex-75 and the Cortex-A55 at Computex, a computer expo held annually in Taipei, Taiwan. The Cortex-A75 will be aimed at the premium segment while the Cortex-A55 core will cater to the mid-range category.

With improved performance, efficiency, and also improvements in machine learning, ARM claims the Cortex-A75 processor offers 50 percent improvement in performance and multi-core capabilities compared to the A73 processor. It can be used in smartphones, “intelligent home devices like DTV, servers, and automotive applications,” says the company.

Jumping to the Cortex-A55, company’s most efficient processor in the mid-range series offers 2.5x improved performance efficiency compared to the current Cortex-A53 based devices.

The latest processors are the first to be built on ARM’s DynamicIQ platform, a more flexible and scalable solution to heterogeneous computing. DynamicIQ allows users greater freedom in choosing the cores — including a 1+7 configuration where a single Cortex-A75 core is paired to seven A55 cores. With these processors, Chipset makers can also use a 4+4, 2+6, 1+3, or other configurations in a single cluster, giving them the expertise to maximise the performance or deliver a SoC that’s equipped for efficiency.

Ditching the cloud-centric approach for AI to more a real-time on device experience, ensuring better privacy and security around AI, the company is expecting a “50x improvement in AI performance over the next three to five years thanks to better architecture, micro-architecture, and software optimisations.”

The company also announced Mali-G72 GPU, with 32 shader cores, 25 percent higher energy efficiency, and a 20 percent better performance density. Accelerating inference engines rather than training engines, Mali-G72 GPU is also optimised to provide an efficient performance for Machine Learning, and mobile VR experiences as well.

As per the company, we should see all these latest designs on mobile devices by the first quarter of 2018.