Alif Semiconductor Unveils Benchmarks for GenAI-Enabled Edge AI MCUs in Ensemble E4, E6, E8 Series

Fast emerging MCU leader Alif Semiconductor has released benchmark results for its Ensemble E4, E6, and E8 microcontrollers and fusion processors.

The Ensemble E4, E6, and E8 devices feature an accelerator for edge AI, the Arm Ethos-U85 neural processing unit, an integrated image signal processor, and a wide memory bus to support image capturing and buffering. These microcontrollers include hardware acceleration for transformer networks, which requires 36 milliwatts for small language model execution. The hardware acceleration supports imaging applications, with object detection completed in less than 2 milliseconds and image classification in less than 8 milliseconds.

Alif Semiconductor introduced neural processing unit-enabled microcontrollers in 2021. The latest Ensemble series devices perform generative AI workloads locally and include features for data processing and imaging tasks. The devices support up to two MIPI-CSI image sensors and a hardware-accelerated image signal processor pipeline that operates at up to 60 frames per second at 2-megapixel resolution. A wide memory subsystem enables on-chip and off-chip transactions, allowing inferencing speeds below one millisecond when running AI models from internal magnetoresistive random-access memory.

The generative AI capability in the Ensemble E4, E6, and E8 devices uses a system architecture. Alif...