As with the previous generation GAP8, Greenwaves is targeting applications at the very edge of the network, such as machine learning in sensors, and other battery operated systems.
The GAP9 includes nine RISC-V cores as well as support for AI and some specific additions to the architecture to support audio processing including music playback, active noise cancellation and voice commands, and freeing up power to improve audio quality through neural network-based features including deep noise reduction and acoustic scene detection.
The GAP9 uses adaptive body bias (ABB) and embedded MRAM, which are available in Globalfoundries 22nm FDSOI manufacturing process and this enables GAP9 to host neural networks with a power efficiency of 330microwatts per GOP. The use of embedded MRAM to reduce by 3.5 times the energy used for transfers of neural network parameters.
The GAP9 also includes the Smart Filter Unit (SFU), a streaming processor that can be used to set up programmable filters for active noise cancellation, spatial sound and other music and voice streaming applications. By taking such AI-adaptive filters away from the general-purpose RISC-V array latencies can be reduced to microseconds.
"Processing a music stream with active noise cancellation consumes less than 10 percent of GAP9's resources, leaving plenty of headroom for dramatically improving the audio comfort or other new cutting-edge features, at the same time simplifying their development. GF’s 22FDX platform, with its incredible performance, ultra-low power capability, low-leakage, and flexibility, was instrumental for enabling us to hit our performance targets," said Loic Lietar, CEO of GreenWaves, in a statement.
The GAP9 is due to go into volume production in summer 2021in either a 3.4mm by 3.4mm by 0.4mm WL-CSP package or a 5.5mm by 5.5mm by 0.5mm ball grid array.
Related links and articles: