Arm launches AI-optimized chips for Android edge biometrics as Sony plans processor integration
The new chips are the Cortex-A77 central processing unit, the Mali-G77 graphics processing unit, and a new Arm Machine-Learning Processor.
The Cortex-A77 provides 20 percent more single-thread processing power, and performs floating-point operations 35 percent faster than its predecessor, which powers the Samsung Galaxy S10. The upgrades improve machine learning by up to 35 times compared to the same chip line two years ago, the company says.
“ML use cases on mobile devices are becoming more complex, so having a CPU that can support this increasing compute demand is vital,” explains Arm Product Management Director for the Cortex-A series Stefan Rosinger. “Just some of these use cases on devices include AI cameras, visual scene detection, 3D scanning, biometric user ID (face recognition), voice recognition, ML in gaming and ML in AR.”
The Mali-G77 is expected to deliver 40 percent higher peak performance than its previous generation, which is also featured in the Samsung flagship smartphone. It consumes 30 percent less power, and runs AI apps up to 60 percent faster, enabling developers to implement more machine learning at the edge.
When the data center server does get involved, the second-generation Machine Learning Processor provides a more scalable design, so manufacturing partners can build 8 processing cores onto a chip capable of 32 trillion operations per second, up from 4.6 trillion previously. The first-generation version was billed as a effective for faster facial recognition when it was launched in early 2018.
Sony AI integration plans
In an overview of Sony’s semiconductor business reported by Ubergizmo, the company has revealed plans to integrate edge AI processing with sensor hardware and pursue a recurring revenue model. The company sees potential for better picture quality in the integration, but also “unlimited possibilities” in recognition and machine vision.
Such integrated technology could be implemented in all kinds of cameras, including smartphone cameras, as well as automobiles, IoT devices, and industrial applications. The company report notes growth in the market for mobile-use image sensors and mobile 3D sensing, and says Sony will pursue software partnerships.
Edge AI capabilities have been integrated into a growing range of camera hardware, including processors launched by Ambarella in January, and programmable chips from Lattice Semiconductor launched earlier this month.