Measure some shit with GPU-Z. Says my average power usage under load is ~151w. Measure with Kill-a-watt. Average is ~200w once I take average system idle off. Both methods are pretty worthless for measuring power usage so I don't know why GPU-Z bothers. The only way to accurately measure is connecting a multimeter with average function to a PCIe riser (Not the PSU rails).
>>55861620
GPU-Z only reports core power usage. It doesn't account for the rest of the card.
>>55861620
Nice!
>>55861620
A few things:
VDDC is for the GPU only, without measuring either converter efficiency losses or VRAM power requirements
An at the wall measurement needs to factor out PSU efficiency losses
Even a "GPU only" benchmark/stress test is going to make the CPU, board (Via PCI-e traffic) and RAM use more power than idle.
>>55861866
The kill-a-watt method is no good because your entire system usage goes up under load. So I would guess that ~200w difference is also additional load on the CPU, other components and the PSU itself which I guess would add ~50w. All are pretty much ball park figures.
Hmm. This is interesting. Toms Hardware measures using the PCIe riser method. I have this particular GPU (image related) and his average usage figure falls more in line with my kill-a-watt measurements. My minimum was ~260w under load using Unigine Valley and my max was ~330w. I need to keep that ~50w variation in mind when looking at GPU-Z in future.
>>55862010
Why are you worrying about this anyway? Considering PSU choices?