The new ARM AI Chips: the development of which lies at the core of Qualcomm, HUAWEI, and other manufacturers, introduced two new chips: the Cortex-M55 and Ethos-U55. They are intended for machine learning and building AI platforms chips based on small compact devices. One of the key features was a markedly increased performance compared to previous solutions.
The ARM Cortex-M55 is the latest model in the Cortex-M line. The vendor claims to increase machine learning productivity by 15 times, and digital signal processing – by 5 times compared to the previous generation of chips due to the use of advanced technologies.
- Qualcomm official video demonstrates the power of flagships with Snapdragon 865
- TechInsider #12: OPPO smartwatches, Xiaomi Mi Mix Alpha 2020, Google Pixel 5 XL
- Smartphones Samsung Galaxy Z Flip vs Motorola Razr: in a real battle
- Smartphone Google Pixel 5 XL (2020) with unusual design and characteristics revealed
Ethos-U55 is a neural processor designed to work in tandem with the Cortex-M55 to further enhance overall performance in demanding scenarios. In this case, the declared increase should provide a 32-fold improvement in machine learning processing compared to the base Cortex-M55, which is 480 times better than in previous generations of Cortex-M chips.
As before, ARM offers only technology, while the partners of the company will be engaged in production. It is stated that new chips require very little energy to work. For example, sensors based on such chipsets can work for several years on a small battery and connect to the Internet only if necessary. It will also increase the security and reliability of user confidential data.
The new platform will allow new-generation devices not only to recognize vibrations but also to detect objects. An example is a 360-degree camera in a cane that can detect obstacles or new sensors for trains that can locally analyze problems to avoid delays. By combining chips, you can get even more impressive features for recognizing gestures and speech.
At the moment, developers have already gained access to the documentation. The release of the first devices based on new chips is scheduled for 2021.