GreenWaves Technologies, a fabless semiconductor vendor focused on ultra-low power edge-based AI processing, recently announced a new member of its GAP IoT application processor family, the GAP9. This latest member combines architectural enhancements using Global Foundries 22-nm FDX process to deliver a peak cluster memory bandwidth of 41.6 Gbytes/s and up to 50 GOPS combined compute power at an overall power consumption of 50 mW.

GAP9 lets OEMs embed machine learning and signal processing capabilities into battery-powered or energy-harvesting devices. This could include IoT sensors in smart buildings or medical wearable devices.

Building on the company’s GAP8 architecture, GAP9 adds support for floating-point arithmetic across all cores based on a transprecision floating-point unit capable of handling floating-point numbers in 8-, 16-, and 32-bit precision with support for vectorization. GAP9 also extends support for fixed-point arithmetic using vectorized 4-bit and 2-bit operations. This positions GAP9 for applications that can take advantage of deep levels of quantization.

GAP9 incorporates bi-directional multichannel, synchronized digital audio interfaces, suiting it for wearable audio products. It also incorporates both CSI-2 and parallel camera interfaces. Other enhancements include the GAP AutoTiler, allowing automatic code generation for neural network graphs and GAPFlow, a series of tools for automating the conversion of neural networks from training packages such as Google TensorFlow.