CAS Reports Sub-10-Millisecond In-Memory Chip

Single-source brief: CAS reports a sub-10-millisecond neural dynamical system chip result. Independent confirmation was not supplied.

What changed

The Chinese Academy of Sciences reported an in-memory computing chip for neural dynamical systems.

CAS links neural dynamical systems to modeling changing physical shapes.

Such work needs dense deformation fields that update in real time.

CAS says current neural dynamical systems can face delays of several hundred milliseconds.

It reports sub-10-millisecond operation for the new chip research.

Technical approach

CAS says the design uses phase-change memory.

It describes this memory as fast, low power, and compatible with CMOS chip processes.

The memory can store several levels of data.

Its electrical conductance can also drift over time.

CAS presents both traits as useful computing features for neural dynamical systems.

The institute says they can support dense storage and parallel computing while maintaining high precision.

Robotics relevance and limits

The reported work could matter for robotics tasks that require fast models of changing shapes.

However, CAS did not report a robot test or an operational deployment.

The supplied evidence gives no chip size, power result, fabrication details, or benchmark method.

It also does not identify the specific task behind the sub-10-millisecond result.

No commercial partner, production plan, or independent validation was provided.