Salon E
Early Power Prediction in Chip Design (303369)
*Sushma Honnavara Prasad, GoogleKeywords: Power, VLSI Design, Data Analysis
Power consumption has received renewed attention in the recent past due to the boom of handheld and mobile electronic devices as well as high-end compute servers. Power has a direct impact on the performance as well as reliability of an electronic device, and hence is one of the most important parameters needed for designing an electronic system. While under-provisioning power delivery or thermal cooling can have an adverse impact on product feasibility, over-provisioning can translate to increased cost or an unwieldy form factor. However, making reliable early power predictions of electronic systems have been challenging. Chip power is influenced by different variables such as voltage, silicon process, temperature, workload and physical layout of the transistors. Overall, this can be viewed as a stochastic process, a collection of random variables representing the evolution of some system of random values over time. Hence, we could benefit from sound statistical and data science principles to address the issue of power prediction.