The Next Step in Low Energy Edge DSP and AI
summary
GAP9 is a RISC‑V AI/DSP chip combining an MCU, a 9‑core compute cluster, a dedicated smart filtering unit for sample‑by‑sample audio, and an NE16 neural accelerator. Measured on an LC3 codec it uses two times fewer cycles and three times less power per operation (≈6× energy efficiency) versus a classic DSP approach. Tooling includes GCC, OpenMP‑style parallelism, PyTorch/TensorFlow flows, a cycle‑accurate SoC simulator and profiler; demos show 48‑channel audio, 1.5 MB RAM targets and battery‑ or harvestable‑powered sensors that run for years.
AI-written summary · about
chapters (11)
- 0:00Introduction to GreenWaves
- 1:23Product lineup and markets
- 3:48Gap9 architecture overview
- 8:23Smart filtering unit deep dive
- 9:26Gap9 vs classic DSP
- 11:35Real-world performance gains
- 12:18Development tools and SDK
- 13:22Customer use cases overview
- 14:47Audio and spatial sound demos
- 16:09AI-driven noise reduction
- 17:13Keyword spotting innovations
related talks
21mNext Generation of GAP8 - IoT App Processor for Inference at the Very Edge
21mGreenWaves Presentation
03mGreenWaves Technologies
20mLow Power AI Edge Hardware
16mRISC-V DSP (P) Extension Proposal
11mRISC-V enabled, low-power CNN classification in Edge devices
30mFrom Eyewear to Silicon: RISC-V for Low-Power AI in Next-Generation Smart Glasses
06mEmbedded Executive LIVE