Imec and GLOBALFOUNDRIES have announced a hardware demonstration of an AI IC made with Imec’s Analog in Memory Computing (AiMC) architecture Using GloFo’s 22nm FD-SOI process.

The chip performs neural network calculations on in-memory computing hardware in the analogue domain.

Achieving record-high energy efficiency up to 2,900 TOPS/W, the accelerator is a key enabler for inference-on-the-edge for low-power devices.

The privacy, security and latency benefits of this new technology will have an impact on AI applications in a wide range of edge devices, from smart speakers to self-driving vehicles.

Neural nets – which depend on large vector matrix multiplications – suffer from the processor-memory bottleneck. However, neural nets can also achieve accurate results if the vector-matrix multiplications are performed with a lower precision on analog technology.

Imec eliminated the bottleneck by performing analogUe computation in SRAM cells.

The resulting Analog Inference Accelerator (AnIA), built on GF’s 22FDX semiconductor platform, has exceptional energy efficiency.

Characterisation tests demonstrate power efficiency peaking at 2,900 tera operations per second per watt (TOPS/W).

Pattern recognition in tiny sensors and low-power edge devices, which is typically powered by machine learning in data centers, can now be performed locally on this power-efficient accelerator.

“The successful tape-out of AnIA marks an important step forward toward validation of Analog in Memory Computing (AiMC),” says Imec’s Diederik Verkest, “the reference implementation not only shows that analogue in-memory calculations are possible in practice, but also that they achieve an energy efficiency ten to hundred times better than digital accelerators. In imec’s machine learning program, we tune existing and emerging memory devices to optimize them for analogue in-memory computation. These promising results encourage us to further develop this technology, with the ambition to evolve towards 10,000 TOPS/W”.

GloFo will include AiMC as a feature able to be implemented on the 22FDX platform for a differentiated solution in the AI market.

Its 22nm FD-SOI process can operate at 0.5 Volt ultralow power and at 1 pico amp per micron for ultralow standby leakage.

22FD-SOI with the AiMC feature is in development at GloFo’s Fab 1 in Dresden.