Ma, XiaoyangXiaoyangMaDeng, ShanShanDengWu, JuejianJuejianWuZhao, ZijianZijianZhaoLehninger, DavidDavidLehningerAli, TarekTarekAliSeidel, KonradKonradSeidelDe, SouravSouravDeHe, XiyuXiyuHeChen, YimingYimingChenYang, HuazhongHuazhongYangNarayanan, VijaykrishnanVijaykrishnanNarayananDatta, SumanSumanDattaKämpfe, ThomasThomasKämpfeLuo, QingQingLuoNi, KaiKaiNiLi, XueqingXueqingLi2023-10-262023-10-262023https://publica.fraunhofer.de/handle/publica/45226110.1109/LED.2023.32743622-s2.0-85159832202This letter proposes C2FeRAM, a 2T2C/cell ferroelectric compute-in-memory (CiM) scheme for energy-efficient and high-reliability edge inference and transfer learning. With certain area overhead, C2FeRAM achieves the following highlights: (i) compared with FeFET/FeMFET, it achieves disturb-free CiM and much higher write endurance (equal to FeRAM), leading to 100× inference time with < 1% accuracy drop for VGG8 in CIFAR-10 dataset, along with the enhanced endurance for weight updates, e.g., CiM-based transfer learning; (ii) compared with 1T1C FeRAM inference cache, the achieved disturb-free feature and CiM capability in C2FeRAM lead to improvements of 4× energy, 200× speed, and 3.2e 5× life cycles. Such benefits highlight an intriguing solution for future intelligent edge AI.encompute-in-memory (CiM)enduranceFeFETFeRAMFerroelectric memoriesread disturbjournal article