Joel Cambonie, co-founder and vice president of engineering at GreenWaves, told eeNews Europe the GAP8 will be twice as energy efficient as cores from ARM ranging from the Cortex-M0 up to the Cortex-M7.
The GAP8 is based on the RISC-V opens-source hardware PULP core developed at the Universities of Bologna and ETF Zurich (see Swiss open-source processor core ready for IoT). It will include eight such cores plus a TensorFlow processing unit (TPU) to accelerate convolutional neural networks for applications such as hardware-based pattern matching.
The processor can be programmed using C/C++/OpenMP compiled using a standard GCC compiler with extensions, Cambonie said.
The chip is targeting TSMC’s 55nm 55LP manufacturing process and is due to tape out in December, Cambonie said, speaking at the Semicon Europa exhibition. Silicon is expected back in February 2017. Simulation predicts that the chip will be able to provide up to 12GOPS of raw performance. For sensor hub type applications the benchmarks show 400MOPS of performance at 1mW of power consumption or 40MOPS at 300 microwatts.
The processor is intended to address such applications as image analytics, motion and vibration analysis and voice recognition and identification amongst others. It will also be able to act as the platform for soft modems to support multiple IoT wireless communications schemes.
Support is included for 802.15.4g, LTE Cat-M and Cat-N1 and 802.11ah (WiFi HaLow). In addition the GreenOFDM software runs on GAP8 and it is claimed reduces the power consumption of OFDM transmissions by an order of magnitude (see Startup’s IoT SoC Packs a Punch).