Over the past decade, emerging spintronics and nanomagnetic devices have attracted a lot of attention due to their versatility, scalability, and energy efficiency. Part of the excitement stems from the discovery and experimental demonstration of spin-transfer-torque (STT) effect , . Together with the tunneling magnetoresistance effect, they provide the means of write and read operations for memory applications. Owing to the nonvolatility of nanomagnets enabling persistent binary states, STT-based magnetic random access memories (STT-MRAMs) can be used both as working memory and for code storage purposes. Most major semiconductor companies are developing STT-MRAM for embedded or standalone applications. A fundamental issue accompanying magnetic switching is its susceptibility to thermal noise. At room temperature, the magnetic switching under STT reacts to thermal fluctuations and often results in a distribution of switching currents and delays. Some proposed applications even explicitly utilize these thermal fluctuations, such as random number generators in spin dice  or stochastic simulation of neuromorphic behavior . In the case of write operation in STT-MRAM, increasing the write-voltage and/or pulsewidth can reduce write error rate (WER) but both quantities are limited by reliability, endurance, and overall performance metrics.