ARM unveiled the next generation of ML and OD processors
The new processors for mobile devices from ARM open the next era in digital technology. The ML processor is sharpened for synchronization and work with machine learning technologies, and the OD processor is a new architecture for object identification.
Today, artificial intelligence technologies are a priority area in the development of many modern devices. As of early 2018, there are only two single-chip mobile processors that have blocks for artificial intelligence – the A11 processor from the big-name Cupertino company and Samsung’s flagship Kirin 970 processor.
These processors are intended exclusively for proprietary companies and are not supplied to the open market.
Solutions from ARM will be freely available and can be used for a variety of modern devices.
Features of the ML processor from ARM
The company has developed this processor completely from scratch using a new proprietary architecture that will quickly and efficiently solve machine learning issues. The processor features excellent performance with up to 4.6 trillion operations per second. In addition to the high performance level, the processor has also received impressive power efficiency parameters, power consumption of the processor at maximum load is 1.5 Watts.
This solution can be used in smartphones, digital cameras, VR devices, drones, medical equipment and even consumer electronics. This universal processor can be easily integrated into new devices and act as an AI coprocessor.
Features of the OD processor from ARM
This processor is designed specifically for devices that are equipped with a camera and are designed to quickly and accurately identify people by facial features and identification of inanimate objects.
The system will be able to identify objects and faces in real time during the video recording or photographing. Moreover, the processor will be able to identify people by their faces at a relatively low quality Full HD video recording. Such parameters mean that the OD processor will be compatible even with budget segment smartphones.
The company claims that their development is 80 times more effective than all currently available facial recognition technologies.
This processor will be able to determine not only the faces, if we talk about the identification of people, but also any body parts, characteristic gestures, the direction of movements and other parameters.
In addition, the manufacturer claims that the joint integration of ML and OD processors in devices will increase the efficiency of their work by several times.