Given that most cards will boost all the way to their highest potential, the numbers we see here should be a good indicator of what to expect from these cards in general. Results are captured via inline power metering instruments primarily Powenetics v2 by Cybenetics/Hardware Busters for accurate card-only data capture. Occassionally, we use PCAT v2 (or v1) for FrameView-related functionalities. Powenetics v2 is the modern iteration of our original Powenetics system licensed from Cybenetics Labs and allows more thorough read-outs for the entire system including the CPU power, GPU power (+PCIe slot) and the system power overall. This removes the guess work from taking power readings from differing GPU APIs and provides consistent readings regardless of vendors.
Clock Speed Behavior and Temp-Power Behavior
Clock speed is directly affected by the GPU and the dynamic boost clock will take advantage of cooler GPUs as it pushes clock speed higher. In the charts below we show the test card’s behavior under gaming load: one chart shows the GPU clock speed and its relation with GPU core temperature while the other shows temperature and relative to power draw.
To interpret the charts, as the card heats up, there is a potential for the GPU speed to go down. A steady decline means that the card cannot sustain its boost clocks which means cooling is inadequate. A consistent temperature with consistent clock speed means this is our stable GPU clock speed and is the intended speed for this GPU.
Contributing to increased heat is power draw. If a GPU cannot fully contain its heat dissipation, higher power draw can cause it to heat-up further. A steady rise in temperatures will result in a decline in clock speeds as well as power draw. The power readings are strictly from the GPU alone as gathered by our instrumentation as indicated above.