FeCAM: A Universal Compact Digital and Analog Content Addressable Memory Using Ferroelectric

Abstract

Ferroelectric field effect transistors (FeFETs) are being actively investigated with the potential for in-memory computing (IMC) over other non-volatile memories (NVMs). Content Addressable Memories (CAMs) are a form of IMC that performs parallel searches for matched entries over a memory array for a given input query. CAMs are widely used for data-centric applications that involve pattern matching and search functionality. To accommodate the ever expanding data, it is attractive to resort to analog CAM for memory density improvement. However, the digital CAM design nowadays based on standard CMOS or emerging nonvolatile memories (e.g., resistive storage devices) is already challenging due to area, power, and cost penalties. Thus, it can be extremely expensive to achieve analog CAM with those technologies due to added cell components. As such, we propose, for the first time, a universal compact FeFET based CAM design, FeCAM, with search and storage functionality enabled in digital and analog domain simultaneously. By exploiting the multi-level-cell (MLC) states of FeFET, FeCAM can store and search inputs in either digital or analog domain. We perform a device-circuit co-design of the proposed FeCAM and validate its functionality and performance using an experimentally calibrated FeFET model. Circuit level simulation results demonstrate that FeCAM can either store continuous matching ranges or encode 3-bit data in a single CAM cell. When compared with the existing digital CMOS based CAM approaches, FeCAM is found to improve both memory density by 22.4× and energy saving by 8.6/3.2× for analog/digital modes, respectively. In the CAM-related application, our evaluations show that FeCAM can achieve 60.5×/23.1× saving in area/search energy compared with conventional CMOS based CAMs.

Countering variations and thermal effects for accurate optical neural networks

“…If MZIs with signicant temperature dierence are physically close to each other, their temperatures also drift due to thermal imbalance. For example, in Figure 5, 7 and 8 have large phases and thus high temperatures, though other MZIs have small phases due to the appended regularization term in (13). These MZIs with small phases can be heated by 7 and 8 and their phases drift accordingly.…”

“…Deep neural networks have achieved remarkable breakthroughs in various classication tasks in recent years. To accelerate the operations in such networks, various hardware platforms with emerging devices, e.g, RRAM/memristor [1][2][3][4][5][6], optical device [7][8][9], spintronics device [10,11] and Ferroelectric Field-Eect Transistor [12][13][14][15] have been proposed. One such promising platform is optical neural networks (ONNs).…”

Deep Random Forest with Ferroelectric Analog Content Addressable Memory

“…However, it is also possible to leverage the analog states of nonvolatile memories for multi-bit or ACAMs [19,20,21]. Multi-bit information can be stored in the CAM, and an analog or multi-bit query can be searched across the CAM array for pattern matching, thus greatly improving the information density and expanding the CAM functionality [19,20,21,22]. In this work, we demonstrate ferroelectric ACAMs and leverage their unique properties to accelerate DRF.…”

“…All the developments above have only considered digital CAMs, where binary information is stored and searched. However, it is also possible to leverage the analog states of nonvolatile memories for multi-bit or ACAMs [19,20,21]. Multi-bit information can be stored in the CAM, and an analog or multi-bit query can be searched across the CAM array for pattern matching, thus greatly improving the information density and expanding the CAM functionality [19,20,21,22].…”

“…In this work, we demonstrate ferroelectric ACAMs and leverage their unique properties to accelerate DRF. In an ACAM cell, a matching range, defined by the upper and lower bounds of the search line (SL) voltage, can be dynamically adjusted by configuring the memory device states [19]. We observe that by fixing the upper/lower bound of the matching range to the maximum/minimum voltage allowed on the SL and leaving the corresponding lower/upper bound adjustable, the respective greater-than (i.e., >)/less-than (i.e., <) branch-split operations in a decision tree can be efficiently implemented in an ACAM cell through a simple search operation, as shown in Fig.…”