ARM launches Cortex-M55 and Ethos-U55 artificial intelligence processors at the same time, both designed for smart speaker devices, cameras, wearables and IoT devices.
The common point of both of these chips is Machine Learning. Cortex-M55 is the first chip to support ARM Helium Vector processing technology (considered part of SIMD Instruction), enabling faster machine learning. Up to 5 times the size, and if it is digital transcription (such as Dolby Atmos), it can work 5 times faster than before.
Cortex-M55 chip will be able to distinguish Intercept user speech (such as OK Google, Hey Siri) or track user location on camera (Such as intelligent surveillance cameras) and if you want to work better than before Smarter than before, it can work with Ethos-U55, a NPU, to speed up this type of processing specifically. The manufacturer can choose how many processors to use. Available in 32 – 256 sets together.
Ethos-U55 can improve machine learning efficiency by 32 times when multiplied by Cortex-M55 that is 15 times higher, meaning that if you use two together, it can increase the ML processing power by 480 times. (This figure is compared with previous Cortex-M models.
When pairing Cortex-M55 with Ethos-U55, the device can use Speech recognition, Gesture detection without having to rely on the cloud for processing. Makes it possible to use even without internet The chip therefore consumes less power than before. Which is a hallmark of Cortex-M series chips
Next year, ARM will allow partners to add Instruction to M55 processors for specialized applications. It is expected that devices with ARM Cortex-M55 and Ethos-U55 chips will be sold for several years. Because this platform is not as competitive as Cortex-A for mobile.