According to fortune reports, Arm introduced the latest Cortex-M processor (M55) and Arm Ethos-U55 miniature neural processing unit (NPU) on Monday.

Like the previous generation, the new Cortex-M55 is Arm's embedded processor. To date, Arm's partners have produced more than 50 billion chips based on Cortex-M designs. The new processors are more powerful and more power efficient, but Arm's main emphasis is on the machine learning capabilities of the chip. It is understood that the M55 is the first processor based on Arm Helium technology for accelerated vector calculations, and runs ML models 15 times faster than the previous generation.

In the past, such chips often lacked enough computing power to effectively perform machine learning functions. Instead, most of these tasks must be completed on high-power chips, such as Arm's Cortex-A series, which is carried by most smartphones in the world.

The Arm Ethos-U55 NPU is designed to accelerate machine learning, while the U55 design will be more streamlined and only work with newer Cortex-M processors such as M55, M33, M7 and M4. Arm said that by running these two chips in combination, the machine learning task can run 480 times faster than the Cortex M chip used in the benchmark test. (The first 15-fold speed increase comes from the M55, and the combination with the Ethos-U55 brings a 32-fold additional increase.) Using these two chips at the same time can also increase energy efficiency by 25-fold, which is important for many battery-powered Equipment is crucial.

Dipti Vachani, Arm's senior vice president and general manager of the automotive and IoT business, said enabling artificial intelligence to run on relatively low-power devices rather than having to maintain continuous communication with cloud-based data centers is critical to data security and privacy. important. Currently, most AI workloads run in these data centers.

She also said that enabling artificial intelligence to work on non-connected, relatively low-power devices is critical to making connected cars, enabling self-driving cars, and introducing machine learning to medical devices.

Arm said that the M55 itself will be able to undertake machine learning tasks ranging from very simple vibration detection (even the previous generation Cortex-M) to target detection in images. When combined with Ethos U55, it can perform higher-level tasks, such as detecting specific gestures, determining whether your fingerprint or facial features match the biometrics already stored on the device, or even voice recognition. However, more computationally intensive tasks, such as classifying a wide variety of objects, or recognizing faces from videos in real time, still require more power-consuming and expensive chips.