Power Consumption Measurements – Details
All power consumption readings have to do with the graphics card alone and not the entire system’s power consumption. I use Powenetics v2, which allows me to measure almost every system component with high accuracy, including the GPU and the CPU, and with very high poll rates, up to 1000 readings per second, so I can “catch” any nasty power spikes.
I provide below more details about my power consumption measurement procedure.
Idle: The system is idle and at UHD resolution, with no windows or active programs. I let the graphics card warm up for more than 15 minutes and I take the average reading over a ten-minute period.
Multi-monitor: I connect two monitors to the graphics card under test. Both are set at 4K resolution. The system is idle with no active windows, and the refresh rate is at 60Hz on both monitors. With higher refresh rates, expect increased power consumption.
Video Playback: I use VLC Media Player to reproduce a 4K 30 FPS video encoded with H.264 AVC with an overall bitrate of 112 MB/s. All modern GPUs support this codec’s decoding.
Gaming: I use Cyberpunk 2077 at UHD for high-end GPUs and QHD for low-end ones that cannot handle this game at UHD or ultra settings, and raytracing is disabled along with DLSS/FSR and Frame Generation. I run my benchmark scene to obtain all required power consumption readings. I also log temperature and GPU fan speed. The latter is used for noise calculations.
Gaming Raytracing: Same as above, but I enable raytracing (Ultra Preset) and disable DLSS, FSR, and Frame Generation.
Furmark: I let this power virus run for 10 minutes to obtain all required power consumption readings. I also log temperature and GPU fan speed. The latter is used for noise calculations.
V-Sync: I run Cyberpunk 2077 at HD resolution at medium graphics settings and with V-Sync enabled. I don’t use high graphics settings to allow all GPUs, even low-end ones, to be close to 60 FPS.
Rendering: I render footage containing many effects in Davinci Resolve at 4K output and with the H.264 codec. I also log temperature and GPU fan speed. The latter is used for noise calculations.
Peak Power Consumption: Before I start Furmark or any other demanding GPU benchmark, the transient load the benchmark applies to the graphics card’s voltage regulator modules leads to power spikes, which I record with the Powenetics v2 system, utilizing its 1000 readings/sec poll rate. These measurements show which PSU is suitable for the specific GPU, but please note that the CPU also has power spikes, so you must also consider them.
Average Power Consumption (GPU only)
System Power Consumption
Peak Power Consumption
