GenAI-Enabled MCUs Empower Edge-Computing Apps
With its focus on next-generation edge AI computing applications, Alif Semiconductor released benchmarks on its latest E4, E6, and E8 microcontrollers and fusion processors, which are able to perform generative AI workloads locally.
The devices, well-suited for high-performance data processing and imaging tasks, can support up to two MIPI-CSI image sensors. They also include a fully hardware-accelerated image signal processor pipeline that operates at up to 60 fps at 2-Mpixel resolution. A wide memory subsystem for fast on- and off-chip transactions pushes inferencing speed well below a millisecond when running AI models from internal MRAM.
Generative AI capability is based on Alif’s system architecture that leverages the Arm Ethos-U85 NPU, which supports transformer-based ML networks. To show its power efficiency, an SLM executed on an E4 device draws only 36 mW of power when generating text to construct a story based on a user-provided prompt.
E4, E6, and E8 devices can implement transformer-based models and generative AI in edge and endpoint products powered by a small battery.
