Physical AI · My own chip
My SoC RISC-V
This chip is a small computer designed to run an inference task without relying on a large processor. It combines a compact RISC-V CPU, an INT8 accelerator and interfaces for external memory and communication. The goal is to bring TinyML to very low-cost hardware.
01
Why it matters
It is a complete design case: reference model, RTL, verification, physical implementation and preparation for open fabrication. It shows that a small team can take AI inference all the way to its own silicon.
02
Specifications
- Type
- RISC-V SoC for TinyML
- Process
- SkyWater SKY130, 130 nm
- CPU
- Bit-serial architecture
- Accelerator
- Signed INT8 MAC
- Boot
- From SPI Flash
- Data memory
- PSRAM
- Demo
- Quantized 4→3 classifier
- Output
- UART
- Verification
- Python reference model, RTL and gate level
- Area
- 6 tiles
- Shuttle
- Headed for Tiny Tapeout TTSKY26d
Your own chip?
Book a free 30-minute callIf your product needs edge inference or custom silicon, let’s talk.See QvIC Labs