FIRESTARTER 2 - dataset and notebooks

FIRESTARTER 2针对最大功耗进行了优化，在两个系统上使用不同的高速缓存访问，每个系统都使用2x英特尔至强E5-2680 v3 (2000 MHz，以避免基于AVX频率的节流)，一个系统使用4x NVIDIA k80。每个值表示超过240 s的平均功率，不包括初始120和最终2 s。每个内存级别都会增加总功耗。每个GPGPU增加29W (空闲) 至156W (应力)。FIRESTARTER 2 is optimized for maximum power consumption. It utilizes distinct cache access patterns across two test systems. Each system is configured with 2x Intel Xeon E5-2680 v3 CPUs clocked at 2000 MHz to avoid AVX-based frequency throttling, and one of the systems is additionally fitted with 4x NVIDIA K80 GPUs. Each reported value represents the average power consumption over a 240-second measurement period, excluding the initial 120 seconds and the final 2 seconds. Each memory level increases the total power consumption. Each GPGPU adds an additional 29 W (idle mode) to 156 W (stressed mode) of power usage.

搜集汇总

背景与挑战

背景概述

该数据集针对最大功耗优化测试，基于英特尔至强E5-2680 v3处理器和NVIDIA k80 GPU系统采集功耗数据。由德累斯顿工业大学于2021年发布，包含相关代码和论文资源。