ARM announced the Ethıs-U55 NPU and Cortex-M55 processor, which it defined as the industry’s first microNPU (Neural Processing Unit). The new processor and NPU duo will provide microcontrollers with serious machine learning capacity.
Machine learning is now almost everywhere and there is a serious competition for the production of NPU (Neural Processing Unit), which is devoted to artificial intelligence with suitable processors. ARM attracted attention with its newly announced processor and NPU duo.
The Cortex-M55 processor announced by ARM and the compatible Ethos-U55 NPU were announced as the industry’s first microNPU, according to the company. The most ambitious area of the company was machine learning. Machine learning performance of NPU increases the performance of microcontrollers by 480 times.
Arm Cortex-M55 processor features
ARM Helium technology is the first time we come across this processor. Thanks to this new technology, the artificial intelligence skills of Cortex-M55 are at the highest level ever. In addition, the processor can use the optimum workload distribution according to the processes being performed.
It is stated that the processor produced with Corstone-300 references will not take too long to enter the market and it can be easily used in any chipset, while it is also quite easy to produce suitable software for Cortex-M55. A single program compatible with many environments has been developed for the processor.
Thanks to its new architecture, the processor can use more variables and process more for 8-bit, 16-bit and 32-bit processes. It also hides critical information from the rest of the apps to protect against software-based attacks.
Features of ARM Ethos-U55
The most striking feature of the NPU was the energy efficiency according to the ARM’s statements. The company announced that NPU saved up to 90% energy in its machine learning tasks. In addition, this NPU will support many different neural networks.
Processes that require intensive processing power will be made directly on the microNPU. It will also ensure that neural networks, activities, and layer aggregation work best while the system is working offline. According to ARM, the memory requirement of the system will decrease by up to 90%.
He states that the new processor can also be used successfully in various artificial intelligence activities such as element usage, classification and definition. The NPU is fully compatible with Cortex M-55 and ARM Corstone systems.