Compute-in-Memory (CIM) Chips
AI, we have a "memory" problem. New computing architectures and chip designs that make more intelligent use of new memory technologies are key to unlocking extreme energy efficiency. We push the frontier of compute-in-memory AI chips through co-designs and prototyping.
Our ongoing research and tapeouts:
- Novel circuit and macro designs for resistive RAM (RRAM), embedded DRAM (eDRAM), and beyond.
- Energy-efficient, precision-configurable, heterogeneous CIM chip for edge transformer acceleration.
- On-chip learning enablement and prototyping.
- We seek new collaborations to apply our unique co-design ideas towards chip prototyping with ferroelectric and magnetic memories!
SAPIENS is the first integrated chip demonstrating on-chip, one-shot learning, enabled by exposing the unique properties of RRAM with the cross-layer design and integration for memory-augmented neural networks. The name for this chip was inspired by how human can learn new concepts with very few or even single example, dating back to the early stage of intelligence about 150 thousand years ago.
Related publications:
VLSI 2021, TED 2021