[17p-M_374-1~8] The Prospects of Data Storage and Non-Volatile Memory Technologies in the AI era

2:00 PM - 2:30 PM JST(5:00 AM - 5:30 AM UTC)

[17p-M_374-2] Trends in Embedded MRAM MCUs for Edge AI Applications

〇Yosuke Tashiro¹, Tomoya Saito¹, Tomoya Ogawa¹, Yasuhiko Taito¹, Takahiro Shimoi¹, Kosei Fujishiro¹, Takashi Ito¹ (1.Renesas)

Keywords:

Edge AI,MCU,MRAM

AI is expanding its use from the cloud to the edge, contributing to improved quality in industrial applications and the realization of smart home systems. To support diverse applications, both MPUs and MCUs must balance performance and power efficiency. With the spread of TinyML, AI inference on MCUs has become increasingly important, but the readout speed of eNVM remains a challenge. To address this, we proposed high-speed random-access MRAM for MCUs and confirmed the advantage by evaluating the impact of weight data locations on inference performance.