Presentation
Improved Peak Power Interval Detection using the Sliding Window Power Profile (SWPP)
DescriptionIn this work, we address the challenge of accurately identifying peak power intervals in the modern SOC design process. The intervals are simulation periods of high-power demand that can impact performance and may lead to chip failure. Existing Interval Based Power Profile (IBPP) methods can be insufficient (up to 50% average power of the interval) for correctly identifying these high average power consuming intervals due to various reasons described in this proposal. To address this issue, jointly with Samsung, we propose a methodology using the Simple Moving Average (SMA) technique to create a Sliding Window Peak Power Profile (SWPP). Our results demonstrate that the SWPP can accurately identify the correct intervals with the highest power consumption, as compared to existing IBPP methods. We apply the SWPP method to a case study involving real Samsung projects and demonstrate its effectiveness compared to existing IBPP methods. Our results suggest that the SWPP is a simple yet effective approach that can be applied to a wide range of other low-power design projects. Jointly developed with Samsung, this method has the potential to improve decision-making in the low-power design process.
Event Type
IP
TimeMonday, July 10th1:45pm - 2:00pm PDT
Location2012, 2nd Floor
IP
RISC-V