IEEE Consumer Electronics Magazine - March/April 2021 - 51

Column: The Art of Storage

MRAM Coming to
Consumer Products
Tom Coughlin
Coughlin Associates

& CONSUMER DEVICES WITH embedded microprocessor-based electronics are an important element in many consumer's lives. These devices
allow us to communicate with each other, augment our senses, monitor our environment, and
help us stay healthy. Memory is an important
element in the successful operation of these
devices, but conventional memory options are
not scaling with demand. NOR flash memory, the
dominant nonvolatile memory in embedded
electronics cannot scale beyond 28 nm and
SRAM, the dominant cache memory takes up a
lot of die space and loses data when power is
removed.
To meet the needs of modern embedded consumer products, the semiconductor foundries
that make these devices are moving to use
emerging nonvolatile technologies that can provide denser memories for AI inference engines
and run for an extended period of time in energy
constrained battery powered conditions.1
TSMC, based in Taiwan, is the largest public
semiconductor foundry. During a recent online
TSMC technology symposium the company spoke
about providing MRAM and RRAM options for
embedded products. The company said that
Digital Object Identifier 10.1109/MCE.2020.3035340
Date of publication 4 November 2020; date of current version
2 February 2021.

March/April 2021

22-nm embedded magnetic random access memory (MRAM) technology IPs are expected to complete reliability qualification in 2020. In addition,
the company's 16 nm MRAM is under development
and is progressing well. Risk production for eFlashlike MRAM (NOR flash replacement) is expected in
Q421 and SRAM-like MRAM is expected in Q422.
The company sees MRAM replacing eFlash in high
reliability MCUs, including for AI, IoT, and automotive AEC-Q100 Grade-1 applications.
At that TSMC event, one of its customers,
Ambiq, announced that it was using MRAM in its
fourth generation Apollo System on Chip (SoC)
family.2 Ambiq is developing chips that will
enable the next generation of battery-powered,
always on, voice recognition IoT endpoint devices. The product sets new standards for ultralow
power intelligent endpoint IoT devices. It is
implemented with TSMC's 22-nm ULL process
and the 32-bit Arm Cortex-M4 core with floating
point unit (FPU) and Arm Artisan physical IP.
The SoC chip uses low-deep sleep current
modes to achieve power use as low as 3 microA/
MHz using MRAM memory and operates up to 192MHz clock frequency with TurboSPOT (subthreshold power-optimized technology), SPOT) and with
a 2D/2.5D graphics accelerator and MIPI DSI 1.2
with up to two lanes at 500 Mb/s per lane.
Figure 1 shows the block diagram of the
Ambiq Apollo4 chip showing various features,
including 2 MB of MRAM along with 1.0 MB of

Published by the IEEE Consumer Technology Society

2162-2248 ß 2020 IEEE

IEEE Consumer Electronics Magazine - March/April 2021

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - March/April 2021