Abstract
With the rapid growth of the internet and embedded systems, large amounts of data are created everywhere in the world simultaneously. Emerging Electronic Applications (ex. Smart Phones, PADs, wireless sensing networks, recognition systems and Internet of things) have characteristics including large amounts of data waiting to be processed, long standby time, and need for low energy. We believe that in the era of big data, energy caused by the large amount of data transfers and complex computations Cloud computing generates could be greatly reduced by introducing a filter like Ternary Content Addressable Memory (TCAM) between different platforms. This kind of memory can compare and filter incoming data, thus reducing the amount of data sent to following stages and save power. However, this kind of filter needs to fulfill the needs of these emerging applications - 1.Large Capacity, 2.long standby time. Traditional Static Random Access Memory (SRAM)-based TCAM needs two pairs of SRAM cells to save 3 states of data (0, 1 and don’t care). Under the same area constraint, SRAM-based TCAM (16T) has a small capacity; In addition, with process scaling down the leakage problem caused by SRAM also appears in TCAM. Conventional solution moves the stored data to another nonvolatile memory macro (NVM macro), but this method not only requires large energy for moving data, but is also a slow process limited by the I/O number of the interface. This study plans to improve current solutions by combining emerging NVM with CMOS process in cell level. Resistive Random Access Memory (ReRAM ) is a very promising nonvolatile memory, with high density, distinctive states ( high R-Ratio), fast random access, and good data retention time. However, a large voltage may be placed across the ReRAM for fast access, which may cause the ReRAM resistance to drift, and could lead to read failure under serious conditions. This situation (Read Disturb) must be put into consideration when designing high-speed ReRAM circuits. This study proposes an RC-filtered stress-decoupled (RCSD) 4T2R nonvolatile TCAM (nvTCAM) to 1) suppress match-line (ML) leakage current from match cells (IML-M), 2) reduce ML parasitic load (CML), 3) decouple NVM-stress from wordlength (WDL) and IML-MIS. RCSD reduces NVM-stress by 6x, and achieves a 4+x improvement in speed-WDL-capacity-product. A 128x32b RCSD nvTCAM macro was fabricated using HfO ReRAM and an 180nm CMOS. This paper presents the first ReRAM-based nvTCAM featuring the shortest (1.2ns) search delay (TSD) among nvTCAMs with WDL≧32bits.